![4 ■ Cognitive Radio Networks
1.3 Noncooperative Sensing Techniques ................................. 9
1.3.1 Energy Detector ............................................... 10
1.3.1.1 Characterization of Energy Detector in AWGN
Channels ............................................. 11
1.3.1.2 Characterization of Energy Detector in Fading
Channels ............................................. 12
1.3.2 Matched Filter Detector ....................................... 12
1.3.2.1 Characterization of the Matched Filter .............. 14
1.3.3 Cyclostationary Feature Detection ............................ 14
1.4 Cooperative Sensing Techniques...................................... 15
1.4.1 Voting-Based Sensing.......................................... 17
1.4.2 Correlator-Based Sensing ...................................... 19
1.4.3 Eigenvalue-Based Sensing ..................................... 20
1.4.3.1 Noise Distribution Unknown, Variance Known .... 23
1.4.3.2 Both Noise Distribution and Variance Unknown... 23
1.5 Conclusions and Open Issues ......................................... 24
References ................................................................... 25
Today, the creation of new radio access technologies is limited by the shortage of the
available radio spectrum. These new technologies are becoming evermore bandwidth
demanding due to their higher rate requirements. Cognitive radio networks and
spectrum-sensing techniques are a natural way to allow these new technologies to be
deployed.
In this chapter, we discuss spectrum sensing for cognitive radio networks. We
begin by introducing the subject in Section 1.1, providing a brief background
followed by a discussion of spectrum-sensing motivations and characteristics. Then
we move on to the spectrum-sensing problem itself in Section 1.2, where we explain
the issues that are inherent to spectrum sensing. In Section 1.3, we explore the
classical noncooperative spectrum-sensing techniques that form the basis for the
more elaborate, cooperative techniques presented in Section 1.4. Finally, we close
this chapter with some conclusions and open issues.
1.1 Introduction
One of the most prominent features of cognitive radio networks will be the ability
to switch between radio access technologies, transmitting in different portions of the
radio spectrum as unused frequency band slots arise [1–3]. This dynamic spectrum
access is one of the fundamental requirements for transmitters to adapt to varying
channel qualities, network congestion, interference, and service requirements. Cog-
nitive radio networks (from now on called secondary networks) will also need to
coexist with legacy ones (hereafter called primary networks), which have the right
to their spectrum slice and thus cannot accept interference.](https://image.slidesharecdn.com/1258189-250522005619-7be5ef49/85/Cognitive-Radio-Networks-Architectures-Protocols-And-Standards-Wireless-Networks-And-Mobile-Communications-1st-Edition-Yan-Zhang-27-320.jpg)
![Spectrum Sensing in Cognitive Radio Networks ■ 5
Based on these facts, underutilization of the current spectrum and the need to
increase the network capacity is pushing research toward new means of exploiting
the wireless medium. In this direction, the Federal Communications Commission
(FCC) Spectrum Policy Task Force has published a report [4] in 2002, in which it
thoroughly investigates the underutilization of the radio spectrum. While the FCC
is in charge of determining the spectrum usage and its policies, the Whitespace
Coalition is studying ways to exploit the spectrum vacancies in the television band.
Cognitive radio networks are envisioned to be able to opportunistically exploit those
spectrum “leftovers,” by means of knowledge of the environment and cognition
capability, to adapt to their radio parameters accordingly. Spectrum sensing is the
technique that will enable cognitive radio networks to achieve this goal.
1.1.1 Interference Management and Spectrum Sensing
To share the spectrum with legacy systems, cognitive radio networks will have to
respect some set of policies defined by regulatory agencies [2,3]. These policies are
based on the central idea where there are primary systems that have the right to the
spectrum and secondary systems that are allowed to use the spectrum so long as they
do not disturb the communications of the primary systems. Roughly speaking, these
policies deal with controlling the amount of interference that the secondary systems
can incur to primary ones. Thus, the problem is one of interference management
[2,3]. We can address this problem from two different points of view: receiver centric
or transmitter centric.
1.1.1.1 Receiver-Centric Interference Management
In the receiver-centric approach [2,3], an interference limit at the receiver is calcu-
lated and used to determine the restriction on the power of the transmitters around
it. This interference limit, called the interference temperature, is chosen to be the
worst interference level that can be accepted without disturbing the receiver operation
beyond its operating point. Although very interesting, this approach requires knowl-
edge of the interference limits of all receivers in a primary system. Such knowledge
depends on many variables, including individual locations, fading situations, mod-
ulations, coding schemes, and services. Receiver-centric interference-management
techniques are not addressed in this chapter as they have been recently ruled out by
the IEEE SCC41 cognitive radio network standard.
1.1.1.2 Transmitter-Centric Interference Management
In the transmitter-centric approach, the focus is shifted to the source of interference
[2,3]. The transmitter does not know the interference temperature, but by means
of sensing, it tries to detect free bandwidth. The sensing procedure allows the
transmitter to classify the channel status to decide whether it can transmit and with](https://image.slidesharecdn.com/1258189-250522005619-7be5ef49/85/Cognitive-Radio-Networks-Architectures-Protocols-And-Standards-Wireless-Networks-And-Mobile-Communications-1st-Edition-Yan-Zhang-28-320.jpg)
![6 ■ Cognitive Radio Networks
how much power. In actual systems, however, as the transmitter does not know the
location of the receivers or their channel conditions, it is not able to infer how much
interference these receivers can tolerate. Thus, spectrum sensing solves the problem
for worst-case scenario, assuming strong interference channels, so that the secondary
system transmits only when it senses an empty medium.
1.1.2 Characteristics of Spectrum Sensing
There are several techniques available for spectrum sensing, each with its own set
of advantages and disadvantages that depend on the specific scenario. Some works
in the literature [5–7] consider spectrum sensing as a method for distinguishing
between two or more different types of signals or technologies in operation. Because
this is not a question of detection (determining whether a given frequency band is
being used), these types of signal identification issues [8] are not addressed in this
chapter. Rather we focus on their detection.
Ultimately, a spectrum-sensing device must be able to give a general picture of the
medium over the entire radio spectrum. This allows the cognitive radio network to
analyze all degrees of freedom (time, frequency, and space) to predict the spectrum
usage. Wideband spectrum-sensing works are also available in the literature [9–12];
however, an equipment able to perform wideband sensing all at once is prohibitively
difficult to build with today’s technology. Feasible spectrum-sensing devices can
quickly sweep the radio spectrum, analyzing one narrowband segment at a time.
This chapter focuses on narrowband-sensing techniques.
In this section, we have emphasized the importance of the spectrum-sensing
technique for cognitive radio networks. In the next section, we aim at understanding
the underlying characteristics of the spectrum-sensing problem, which will enable
us to develop the approaches presented further in this chapter.
1.2 Problem Formulation
1.2.1 The General Spectrum-Sensing Problem
Spectrum sensing is based on a well-known technique called signal detection. In a
nutshell, signal detection can be described as a method for identifying the presence
of a signal in a noisy environment. Signal detection has been thoroughly studied for
radar purposes since the 1950s [13]. Analytically, signal detection can be reduced
to a simple identification problem, formalized as a hypothesis test [14–16]:
y(k) =
n(k): H0
s(k) + n(k): H1
, (1.1)
where
y(k) is the sample to be analyzed at each instant k
n(k) is the noise (not necessarily white Gaussian noise) of variance σ2](https://image.slidesharecdn.com/1258189-250522005619-7be5ef49/85/Cognitive-Radio-Networks-Architectures-Protocols-And-Standards-Wireless-Networks-And-Mobile-Communications-1st-Edition-Yan-Zhang-29-320.jpg)
![Spectrum Sensing in Cognitive Radio Networks ■ 7
H0
H1
H0
H1
P(H0|H0)
P(H0|H1)
P(H1|H0)
P(H1|H1)
Figure 1.1 Hypothesis test and possible outcomes with their corresponding
probabilities.
s(k) is the signal the network wants to detect
H0 and H1 are the noise-only and signal-plus-noise hypotheses, respectively
H0 and H1 are the sensed states for the absence and presence of signal, respec-
tively. Then, as shown in Figure 1.1 we can define four possible cases for the detected
signal:
1. Declaring H0 when H0 is true (H0|H0)
2. Declaring H1 when H1 is true (H1|H1)
3. Declaring H0 when H1 is true (H0|H1)
4. Declaring H1 when H0 is true (H1|H0)
Case 2 is known as a correct detection, whereas cases 3 and 4 are known as
a missed detection and a false alarm, respectively. Clearly, the aim of the signal
detector is to achieve correct detection all of the time, but this can never be perfectly
achieved in practice because of the statistical nature of the problem. Therefore, signal
detectors are designed to operate within prescribed minimum error levels. Missed
detections are the biggest issue for spectrum sensing, as it means possibly interfering
with the primary system. Nevertheless, it is desirable to keep the false alarm rate
as low as possible for spectrum sensing, so that the system can exploit all possible
transmission opportunities.
The performance of the spectrum-sensing technique is usually influenced by
the probability of false alarm Pf = P(H1|H0), because this is the most influential
metric. Usually, the performance is presented by receiver operation characteristic
(ROC) curves, which plot the probability of detection Pd = P(H1|H1) as a function
of the probability of false alarm Pf .
Equation 1.1 shows that, to distinguish H0 and H1, a reliable way to differentiate
signal from noise is required. This becomes very difficult in the case where the
statistics of the noise are not well known or when the signal-to-noise ratio (SNR) is
low, in which case the signal characteristics are buried under the noise, as shown by
Tandra et al. in [17]. In fact, this work also shows that the lesser one knows about](https://image.slidesharecdn.com/1258189-250522005619-7be5ef49/85/Cognitive-Radio-Networks-Architectures-Protocols-And-Standards-Wireless-Networks-And-Mobile-Communications-1st-Edition-Yan-Zhang-30-320.jpg)
![8 ■ Cognitive Radio Networks
the statistics of the noise, the worse the performance of any signal detector is in the
low-SNR regime.
Clearly, the noise characteristics are very important for the spectrum-sensing
procedure. Most works on spectrum sensing consider noise to be additive white
Gaussian noise (AWGN), because many independent sources of noise are added
(central limit theory). Nevertheless, in realistic scenarios, this approximation may
not be appropriate, because receivers modify the noise through processes such as
filters, amplifier nonlinearities, and automatic gain controls [18,19].
Poor performance in a low-SNR regime means that all of the techniques available
are negatively affected by poor channels. In the case of variable channel gains,
Equation 1.1 is rewritten as
y(k) =
n(k): H0
h(k)s(k) + n(k): H1
, (1.2)
where h(k) is the channel gain at each instant k. In a wireless radio network, as it is
reasonable to assume that the spectrum-sensing device does not know the location
of the transmitter, two options arise:
■ A low h(k) is solely due to the pathloss (distance) between the transmitter
and the sensing device, meaning that the latter is out of range.
■ A low h(k) is due to shadowing or multipath, meaning that the sensing device
might be within the range of the transmitter.
In the latter case, a critical issue arises. Therein, the fading plays an especially
negative role in the well-known “hidden node” problem [20]. In this problem, the
spectrum-sensing terminal is deeply faded with respect to the transmitting node
while having a good channel to the receiving node. The spectrum-sensing node then
senses a free medium and initiates its transmission, which produces interference on
the primary transmission. Thus, fading here introduces uncertainty regarding the
estimation problem. To solve this issue, cooperative sensing has been proposed. In
this approach, several sensing terminals gather their information to make a joint
decision about the medium availability. Cooperative spectrum sensing is further
explored in Section 1.4.
1.2.2 Spectrum Sensing from the Cognitive Radio Network
Perspective
In contrast to the general case, where only the signal detection aspect is considered,
the problem of spectrum sensing as seen from a cognitive radio perspective has
very stringent restrictions. These are mainly imposed by the policies these cognitive](https://image.slidesharecdn.com/1258189-250522005619-7be5ef49/85/Cognitive-Radio-Networks-Architectures-Protocols-And-Standards-Wireless-Networks-And-Mobile-Communications-1st-Edition-Yan-Zhang-31-320.jpg)
![Spectrum Sensing in Cognitive Radio Networks ■ 9
radio networks face to be able to operate alongside legacy networks. Some of these
restrictions are summarized in Sections 1.2.2.1 through 1.2.2.3.
1.2.2.1 No Prior Knowledge on the Signal Structure
There are portions of the spectrum where multiple technologies (using different
protocols) share the spectrum, such as the ones operating on the instrumentation
scientific and medical (ISM) unlicensed band. Cognitive radio networks must be able
to deal with existing multiple technologies, as well as new ones that may eventually
appear across the span of the wireless radio spectrum. These networks should be
able to discover the state of the medium irrespective of the technologies in use. Of
course, if the technologies are known, then this information can be exploited to
improve the accuracy of the spectrum sensing, for example, through the detection
of known pilot sequences within the signal [17].
1.2.2.2 Sensing Time
Due to the primary importance of the legacy system, the secondary system must be
designed to free the medium as soon as it senses that a legacy network has initiated
a transmission. For efficient use of the spectrum, these secondary networks must
also sense available spectrum as quickly as possible, in the least possible number
of received samples. In general terms, spectrum-sensing techniques work through a
compromise between the number of samples and accuracy. Cooperative spectrum
sensing gives the opportunity to decrease the sensing time for the same level of
accuracy.
1.2.2.3 Fading Channels
As discussed earlier, spectrum sensing is particularly sensitive to fading environments.
Communication systems operate in diverse environments, including those prone to
fading. Thus, in many situations, spectrum-sensing devices must be able to detect
reliably even over heavily faded channels. Although several works have focused on
sensing for the fading environment in the noncooperative setting [21,22], it is
foreseen that cooperative sensing [23–31] is the best way to address this problem.
Nevertheless, it creates other implications such as the distribution of metrics among
the sensing terminals and the decision regarding which terminals are to be considered
dependable or not.
1.3 Noncooperative Sensing Techniques
In a realistic spectrum-sensing scenario, there are situations in which only one
sensing terminal is available or in which no cooperation is allowed due to the lack](https://image.slidesharecdn.com/1258189-250522005619-7be5ef49/85/Cognitive-Radio-Networks-Architectures-Protocols-And-Standards-Wireless-Networks-And-Mobile-Communications-1st-Edition-Yan-Zhang-32-320.jpg)
![10 ■ Cognitive Radio Networks
of communication between sensing terminals. In this section, we explore the main
single-user sensing schemes, some of which will serve as a basis for the development
of the cooperative ones investigated in Section 1.4.
Single-user spectrum-sensing approaches have been widely studied in the lit-
erature, in part because of the relationship to signal detection. There are several
classical techniques for this purpose, including the energy detector (ED) [16,21,22],
the matched filter (MF) [25,32], and the cyclostationary feature detection (CFD)
[6,33–36].
1.3.1 Energy Detector
The most well-known spectrum-sensing technique is the ED. It is based on the
principle that, at the reception, the energy of the signal to be detected is always
higher than the energy of the noise. The ED is said to be a blind signal detector
because it ignores the structure of the signal. It estimates the presence of a signal by
comparing the energy received with a known threshold ν [16,21,22], derived from
the statistics of the noise.
Let y(k) be a sequence of received samples k ∈ {1, 2, . . . , N} at the signal detector,
such as that in Equation 1.1. Then, the decision rule can be stated as
decide for
H0, if E ν
H1, if E ≥ ν
,
where
E = E[| y(k) |2] is the estimated energy of the received signal
ν is chosen to be the noise variance σ2
In practice, one does not dispose of the actual received energy power E. The ED
uses instead the approximation Ê, where
Ê
1
N
N
k=1
| y(k) |2
.
As the number of samples N becomes large, by the law of the large numbers, Ê
converges to E.
The ED is one of the simplest signal detectors. Its operation is very straightfor-
ward, and it has a very easy implementation, because it depends only on simple and
readily available information.
Nevertheless, in spite of its simplicity, the ED is not a perfect solution. The
approximation of signal energy E gets better as N increases. Thus, the performance
of the ED is directly linked to the number of samples. Furthermore, the ED relies
completely on the variance of the noise σ2, which is taken as a fixed value. This is
generally not true in practice, where the noise floor varies. Essentially, this means](https://image.slidesharecdn.com/1258189-250522005619-7be5ef49/85/Cognitive-Radio-Networks-Architectures-Protocols-And-Standards-Wireless-Networks-And-Mobile-Communications-1st-Edition-Yan-Zhang-33-320.jpg)
![Spectrum Sensing in Cognitive Radio Networks ■ 11
ν
(a) (b)
H0 H1
ν
H0 H1
ε̂ ε̂
σ2
Figure 1.2 (a) Ideal ED scheme. (b) Detection uncertainty for the ED.
that the ED will generate errors during those variations, especially when the SNR is
very low, as shown in Figure 1.2b, where we see an area of uncertainty surrounding
the threshold ν in contrast to the case portrayed in Figure 1.2a, in which perfect
noise knowledge is considered.
1.3.1.1 Characterization of Energy Detector in AWGN Channels
This case has been studied in the work of Urkowitz in 1967 [16]. It is known
that the energy detection is the optimal signal detector in AWGN considering no
prior information on the signal structure [17]. To understand the inner workings
of the ED in this scenario, we need to understand how the probability of detection
Pd = Prob{Ê ν|H1} and false alarm Pf = Prob{Ê ν|H0} behave with the
measured received signal energy.
Take n(k) ∼ NC
0, σ2
as the AWGN noise sample. Then, we know that for
the noise-only case, the distribution of the energy of n over T samples can be
approximated by a zero mean chi-square distribution χ2
2TW [16], where W is the
total bandwidth. Similarly, the energy over T samples of the sum of a signal plus
noise can be represented by a noncentral chi-square distribution χ2
2TW (λ) [16],
where λ is the noncentrality parameter. Briefly:
Ê ∼
χ2
2TW , H0
χ2
2TW (λ), H1
.
With these considerations in mind
Pf = Qm(
λ ξ,
√
ν) (1.3)
and
Pd =
(TW , ν/2)
(TW )
, (1.4)](https://image.slidesharecdn.com/1258189-250522005619-7be5ef49/85/Cognitive-Radio-Networks-Architectures-Protocols-And-Standards-Wireless-Networks-And-Mobile-Communications-1st-Edition-Yan-Zhang-34-320.jpg)
![12 ■ Cognitive Radio Networks
where
Qm is the Marcum Q-function
is the gamma function
ξ is the SNR seen by the signal detector
1.3.1.2 Characterization of Energy Detector in Fading Channels
In 2002, Kostylev [21] studied the performance of the ED in fading channels. He
derived analytical expressions for the ED over the Rayleigh fading channel case (also
analyzed the Rice and Nakagami cases numerically). In 2003, the problem was
revisited by Digham et al. [22], who provided an alternative analytical development
for these three kinds of fading channels. In this section, however, we will restrict the
analysis to the more commonly adopted Rayleigh fading.
Let us begin by recalling that, in this case, the model of interest is the one
shown in Equation 1.14. As such, similar to what Urkowitz did in [16], Kostylev
characterized the statistics of the energy of the signal for both the H0 and H1 cases,
under the assumption that h(k) is Rayleigh distributed:
Ê ∼
χ2
2(TW +1), H0
e2(ξ2+1) + χ2
2TW (λ), H1
,
where
e2(d2+1) is the exponential distribution with parameter α = 2(ξ2 + 1) with
probability density function f (x, α) = αe−αx
ξ is the SNR
It is clear that, under the hypothesis H0, the statistics are the same as for the
AWGN channel case, so the probability of false alarm is the same as in Equation 1.3.
Pf = Qm(
λ ξ,
√
ν). (1.5)
The H1 case behaves differently and has the probability of detection given
by [22]
Pd = e
Ê
2
TW −2
m=0
1
m!
Ê
2
+
1 + ξ
ξ
TW −1
e
Ê
2(1+ξ) − e
Ê
2
TW −2
m=0
1
m!
Êξ
2(1 + ξ)
.
(1.6)
1.3.2 Matched Filter Detector
We have seen previously in Section 1.3.1 that the best sensing technique in an
AWGN environment, and without any knowledge of the signal structure, is the](https://image.slidesharecdn.com/1258189-250522005619-7be5ef49/85/Cognitive-Radio-Networks-Architectures-Protocols-And-Standards-Wireless-Networks-And-Mobile-Communications-1st-Edition-Yan-Zhang-35-320.jpg)
![Spectrum Sensing in Cognitive Radio Networks ■ 13
ED. If we do assume some knowledge of the signal structure, then we can achieve a
better performance.
Most of the wireless technologies in operation include the transmission of some
sort of pilot sequence to allow channel estimation, to beacon its presence to other
terminals, and to give a synchronization reference for subsequent messages. Sec-
ondary systems can exploit pilot signals to detect the presence of transmissions of
primary systems in their vicinity.
If a pilot signal is known, then the MF signal detector achieves the optimal
detection performance in AWGN channel, since it maximizes the SNR, as shown
by Tandra and Sahai in [17].
Let us assume that
■ The signal detector knows the pilot sequence x(k), the bandwidth, and the
center frequency in which it will be transmitted.
■ The pilot sequence is always appended to the transmission of each primary
system (uplink or downlink).
■ The signal detector can always receive coherently.
Then, if y(k) is a sequence of received samples at instant k ∈ {1, 2, . . . , N} at the
signal detector, the decision rule can be stated as [25]
decide for
H0, if Ŝ ν
H1, if Ŝ ≥ ν
,
where
Ŝ =
N
k=1
y(k)x(k)∗
(1.7)
is the decision criterion
ν is the threshold to be compared
x(k)∗ is the transpose conjugate of the pilot sequence
Here the threshold ν is not the noise variance as it was for the ED. The hypothesis
decision is simplified as the MF maximizes the power of Ŝ as shown in Equation 1.7.
This means that it performs well even in a low-SNR regime.
The MF has some drawbacks. First, a cognitive spectrum sensor might not know
which networks are in operation in the environment at a given moment. Therefore,
it may not know which sets of pilots to look for. One must remember that if it
tries to match an incorrect pilot, it will sense an empty medium and will incorrectly
conclude that the medium is free. Second, the MF requires that every medium access
be “signed” by a pilot transmission, but this is not the case in general. Furthermore,
pilot sequences are only transmitted in the downlink direction. This leaves the](https://image.slidesharecdn.com/1258189-250522005619-7be5ef49/85/Cognitive-Radio-Networks-Architectures-Protocols-And-Standards-Wireless-Networks-And-Mobile-Communications-1st-Edition-Yan-Zhang-36-320.jpg)
![14 ■ Cognitive Radio Networks
uplink transmissions uncovered. Third, the MF requires coherent reception, which
is generally hard to achieve in practice.
1.3.2.1 Characterization of the Matched Filter
Signal detection using the MF was studied in 2006 by Cabric et al. in [25]. They
showed that Ŝ is Gaussian:
Ŝ ∼
N
0, σ2
nε
, H0
N
ε, σ2
nε
, H1
,
where σ2
n is the variance of the noise and
ε =
N
k=1
x(k)2
.
Based on this information, the probabilities of false alarm Pf and detection
Pd are
Pf = Q
Ŝ
εσ2
n
(1.8)
and
Pd = Q
Ŝ − ε
εσ2
n
. (1.9)
1.3.3 Cyclostationary Feature Detection
As we have seen, although it performs well, even in the low-SNR regime, the MF
requires a good knowledge of the signal structure, which secondary terminals may
not have. The natural question to ask is whether we can still be able to perform
spectrum sensing with a limited knowledge of the signal structure, perhaps based on
a characteristic that is common to most known transmitted signals. In the following
text, we show that it is indeed possible.
The cyclostationary feature detector relies on the fact that most signals exhibit
periodic features, present in pilots, cyclic prefixes, modulations, carriers, and other
repetitive characteristics [6,33–37]. Because the noise is not periodic, the signal can
be successfully detected.
The works by Gardner [33] in 1991 and Enserink et al. [34] in 1995 have studied
this signal detection scheme in detail. The work of Enserink et al. follows the same
line of the one by Gardner, in which the cyclostationary feature detector is based on](https://image.slidesharecdn.com/1258189-250522005619-7be5ef49/85/Cognitive-Radio-Networks-Architectures-Protocols-And-Standards-Wireless-Networks-And-Mobile-Communications-1st-Edition-Yan-Zhang-37-320.jpg)
![Spectrum Sensing in Cognitive Radio Networks ■ 15
the squared magnitude of the spectral coherence, which for any random process X
is given by
|ρα
X ( f )| =
|Sα
X ( f )|2
SX
f + α
2
SX
f − α
2
1
2
, (1.10)
where
SX is the spectral correlation density function
α is the cyclic frequency
f is the spectral frequency
In the specific case of the cyclostationary feature detector, substituting ρα
X ( f ) by
ρ̂α
X ( f ) and SX by ŜX , which are the estimated versions of the same quantities, we
have the decision metric:
M̂ = |ρ̂α
X ( f )| =
|Sα
X ( f )|2
ŜX
f + α
2
ŜX
f − α
2
, (1.11)
which goes into the decision statistic, given by
decide for
H0, if M̂ ν
H1, if M̂ ≥ ν
,
A recent work focuses on a cyclostationary feature detector for cognitive radio
networks [37], called multi-cycles detector. In this work, a cyclostationarity detector
scheme is employed on nonfiltered signals, such as OFDM, to detect the cyclic
frequency and its harmonics. Finally, it is thought that the cyclostationary feature
detector is the most promising signal detection technique as it combines good
performance with low requirements on the knowledge of the signal structure [35].
1.4 Cooperative Sensing Techniques
Although for simple AWGN channels most classical approaches perform well, as we
have seen, in the case of fading these techniques are not able to provide satisfactory
results due to their inherent limitations and to the hidden node problem. To this end,
several works [23–31] have looked into the case in which cooperation is employed
in sensing the spectrum.
Consider the scenario depicted in Figure 1.3, in which primary users (in white)
communicate with their dedicated (primary) base station. Secondary receivers
{RX1, RX2, RX3, . . . , RXK } cooperatively sense the channel to identify a white
space and exploit the medium. The main idea of the cooperative sensing techniques
is that each receiver RXi can individually measure the channel and interact on their](https://image.slidesharecdn.com/1258189-250522005619-7be5ef49/85/Cognitive-Radio-Networks-Architectures-Protocols-And-Standards-Wireless-Networks-And-Mobile-Communications-1st-Edition-Yan-Zhang-38-320.jpg)
![16 ■ Cognitive Radio Networks
RX1
RX2
RX3
RXK
Figure 1.3 Cooperative sensing scenario.
findings to decide if the medium is available. The main drive behind this idea is
that each secondary receiver will have a different perception of the spectrum, as its
channel to the receiver will be different from the other secondary receivers, thus
decreasing the chances of interfering with hidden nodes.
We will concentrate on the scenario depicted in Figure 1.3, although all sensing
techniques presented herein can be also applied to alternative scenarios available in
the literature, that is, [38].
The cooperative spectrum sensing can be [31]
■ Centralized, in which a central entity gathers all information from all sec-
ondary receivers to make a decision about the medium status, which is then
transmitted back to the receivers
■ Distributed, in which the receivers share their information to make their own
decision
In both these situations, the cooperative spectrum sensing is plagued with one
problem: how to report or distribute the measures in a resource-constrained network.
In fact, if these measurements are the basis for deciding whether a transmission can
be made or not, then it does not make any sense to propagate the measurements
before the decision is made. To overcome the problem, one could create a dedicated
channel for signaling (such as that in [39]) or use an unregulated band (such as
ISM). Other works [23–27,30,31] try to restrict the reporting to the minimum
possible (often one bit) to ease the process of distributing this information. Finally,
[28] considers a hierarchically structured secondary network, in which the secondary
spectrum sensors are the secondary base stations, distributed over the sensing area.
These base stations would make use of a backbone with enough bandwidth to
distribute the measurements among themselves, irrespective of being a single bit or
the actual acquired data. Then, during a white space, the terminals are allowed to
transmit. Nevertheless, secondary base stations, as opposed to secondary terminals,](https://image.slidesharecdn.com/1258189-250522005619-7be5ef49/85/Cognitive-Radio-Networks-Architectures-Protocols-And-Standards-Wireless-Networks-And-Mobile-Communications-1st-Edition-Yan-Zhang-39-320.jpg)
![Spectrum Sensing in Cognitive Radio Networks ■ 17
RX1
RX2
RX3
RX4
Figure 1.4 Cooperative sensing scenario.
have more processing power and fewer power constraints so that they can perform
the spectrum-sensing task better. It should be noted that both of these approaches
have their own target applications; neither can be considered the best approach in
every case.
Another problem of cooperative spectrum sensing is identifying which sec-
ondary receivers offer reliable estimations. Let us consider the situation depicted
in Figure 1.4, in which one primary terminal is transmitting data in the uplink
channel (with low power) toward its primary base station. Several spectrum sensors
{RX1, RX2, RX3, RX4} are monitoring the medium detect its state. In this example,
{RX2, RX3, RX4} are in range of the transmitter and can correctly sense its ongoing
transmission, but RX1 is not.∗ Thus, when the measures of all of the sensors are
gathered, how does one select the individual receivers that are performing a reliable
measurement? Without knowing the position of the primary transmitter and the
channels between secondary receivers and the primary transmitter, this is a com-
plicated task. The work by Mishra et al. [30] looks further into the performance
impacts of the lack of reliability. Some works [35,40] discuss about a weighting
scheme to give different scales to different secondary receivers based on their chan-
nel. Other works [23–27] propose a voting scheme to make a trustworthy decision,
even with the presence of doubtful measurements.
In the remainder of this section, we explore some of the state-of-the-art
cooperative sensing techniques.
1.4.1 Voting-Based Sensing
We saw in Section 1.3.1 that, in the low-SNR regime, the ED is highly vulnerable to
fading and fluctuations in the level of the noise power. What if, instead of employing
∗ This would also apply to the case where RX1 is shadowed or is in a deep multipath fading.](https://image.slidesharecdn.com/1258189-250522005619-7be5ef49/85/Cognitive-Radio-Networks-Architectures-Protocols-And-Standards-Wireless-Networks-And-Mobile-Communications-1st-Edition-Yan-Zhang-40-320.jpg)
![18 ■ Cognitive Radio Networks
the ED at one location, we could do the same thing in other locations as well? It
is expected that among several secondary receivers, even though some will suffer
from fading or imprecisions due to the choice of the threshold, some will be able to
correctly sense the medium. This is the main idea behind the collaborative spectrum
sensing based on voting, studied in a number of works [23–27].
In the voting spectrum sensing, each secondary receiver RXi uses spectrum
sensing to form its own decision, as presented in Section 1.3.1. Consider the vector
of all responses r such that
r = [r1 r2 r3 . . . rK ] ,
where ri ∈ {1, 0} is the binary response for each sensor i. After all measurements are
gathered, the voting procedure takes place [23–25]:
decide for
H0, if V = 0
H1, if V ≥ 0
,
where
V =
K
k=1
rk.
Briefly, the voting schemes select H1 if at least one of the secondary receivers
decides for H1, which is known as the OR rule. Although this may seem too
pessimistic, as it will favor false alarms, according to [23–25], this already gives
improvements over the simple energy detection case even for two users. This is
reasonable if we remark that with a high number of sensors, higher the probability
of reliable spectrum sensing among secondary receivers will be. The probabilities of
detection and false alarm for the cooperative approach are
Qf = 1 − (1 − Pf )K
(1.12)
and
Qd = 1 − (1 − Pd)K
, (1.13)
respectively.
The work by Sun et al. [26] revisits this scheme to estimate the reliability of each
node. In this scheme, only the nodes with reliable sensing are allowed to report their
detection. The reliability measure is based on how close the energy of y(k) is to ν,
as shown in Figure 1.5.
This work defines two new thresholds, ν1 and ν2, that are used to define a “no
decision” region. Thus the decision rule can be stated as
decide for
H0, if 0 ≤ E ≤ ν1
H1, if E ≥ ν2
.](https://image.slidesharecdn.com/1258189-250522005619-7be5ef49/85/Cognitive-Radio-Networks-Architectures-Protocols-And-Standards-Wireless-Networks-And-Mobile-Communications-1st-Edition-Yan-Zhang-41-320.jpg)
![Spectrum Sensing in Cognitive Radio Networks ■ 19
ν
ν1 ν2
H0 H1
E (|y(k)|2)
No detection
Figure 1.5 Reliability decision scheme.
If E falls in (ν1, ν2), then the secondary receiver decides not to report. This way
the overall decision, based on the OR rule, concentrates on the reports of M users
with a reliable detection out of K total users. The results from this work suggest an
increased performance over the conventional case, where no reliability information
is used.
Another work by Sun et al. [27] proposes a cluster-based spectrum sensing. In
this work, a cluster is a grouping of secondary receivers that are spatially close. In
each cluster, one receiver, called the cluster head, is elected to do the local decision
and the reporting to the central decision entity. There the final decision takes place.
1.4.2 Correlator-Based Sensing
Another possibility is to gather all received samples at a central entity that will
take the decision instead of leaving the decision of the medium availability to the
secondary receivers. With an overall view of the situation, the central entity can
decide how to manage the measurements for the decision-taking task better. The
schemes presented in this and the following sections all involve such a central entity.
Let us, for simplicity sake, suppose that all secondary receivers {RX1,
RX2, RX3, . . . , RXK }, shown in Figure 1.3 are within the range of a certain primary
transmitter. Then, considering a flat-faded environment, we have
yi(k) =
ni(k): H0
hi(k)s(k) + ni(k): H1
,
where the subscript i means that each value is to be taken for each user i.
We can see that for the H0 hypothesis, all yi(k) are independent because they are
only composed of AWGN noise. On the other hand, in the H1 hypothesis, all yi(k)
are composed of not only the noise but also the signal component s(k) modulated
by the channel hi(k). As we know, the signal is common for all users, because it is
broadcast by the primary transmitter. We can exploit this fact to detect the presence
of transmitted signals by focusing on the correlation between received signals from
secondary receivers.](https://image.slidesharecdn.com/1258189-250522005619-7be5ef49/85/Cognitive-Radio-Networks-Architectures-Protocols-And-Standards-Wireless-Networks-And-Mobile-Communications-1st-Edition-Yan-Zhang-42-320.jpg)
![20 ■ Cognitive Radio Networks
This correlation is calculated via the cyclic convolution, defined as
R(ij)(k) =
N
k=1
yi(a)yj((k − a) mod N)
where i and j are the indices of any two secondary receivers.
In this scheme, the decision rule is given by
decide for
H0, if L ν
H1, if L ≥ ν
,
where L is the decision statistic calculated as
L = max
(i,j)∈B
max
k
(R(ij)), (1.14)
where
R(ij) is the pairwise cyclic convolution for all permutation of secondary receivers
B = {(x, y) ∈ A × A|y ≥ x + 1}, A = {1, 2, . . . , N}. Note that unlike the MF,
this scheme does not require coherent reception, as it looks for the highest
correlation between any two pairs of sensors. Nevertheless, in the case of
coherent reception, we could rewrite Equation 1.14 as
L =
N
k=1
(R(ij)), ∀ (i, j) | i = j,
which would effectively maximize the SNR.
As far as the authors know, this spectrum sensing scheme has not yet been studied
in the literature and thus its performance is not known. It would likely suffer from
the same problem as the MF, namely, the challenge of correctly choosing ν. The
main limitation of this scheme would be its necessity to report all the measurements,
which would require an infrastructure with a very high bandwidth dedicated for the
task.
1.4.3 Eigenvalue-Based Sensing
Eigenvalue-based sensing is another technique for cooperative sensing, introduced
by Cardoso et al. [28] and Zeng et al. [29], based on evaluating the eigenvalues
of a matrix formed by the samples collected by multiple sensors in relation to the
Marchenko–Pastur law. Herein, we explore the approach as was presented in [28]
because the approach in [29] is very similar.
To better understand how this spectrum-sensing procedure works, we start with
the following assumption:](https://image.slidesharecdn.com/1258189-250522005619-7be5ef49/85/Cognitive-Radio-Networks-Architectures-Protocols-And-Standards-Wireless-Networks-And-Mobile-Communications-1st-Edition-Yan-Zhang-43-320.jpg)
![Spectrum Sensing in Cognitive Radio Networks ■ 21
■ The K base stations in the secondary system share information between them.
This can be performed by transmission over a wired high-speed backbone.
■ The base stations are analyzing the same portion of the spectrum.
Let us consider the following K × N matrix consisting of the samples received
by all the K secondary receivers RXi:
Y =
⎡
⎢
⎢
⎢
⎢
⎢
⎣
y1(1) y1(2) · · · y1(N)
y2(1) y2(2) · · · y2(N)
y3(1) y3(2) · · · y3(N)
.
.
.
.
.
.
.
.
.
yK (1) yK (2) · · · yK (N)
⎤
⎥
⎥
⎥
⎥
⎥
⎦
.
Then, the objective of the eigenvalue-based approach is to perform a test of
independence of the signals received at RXi. As said before, in the H1 case, all
the received samples are expected to be correlated, whereas in the H0 case, the
samples are decorrelated. Hence, in this case, for a fixed K and N → ∞, under the
H0 assumption the sample covariance matrix 1
N YYH converges to σ2I. However,
in practice, N can be of the same order of magnitude as K and therefore one
cannot infer directly 1
N YYH independence of the samples. This can be formalized
using tools from random matrix theory [41]. In the case where the entries of Y are
independent (irrespective of the specific probability distribution, which corresponds
to H0), we can use the following result from asymptotic random matrix theory [41]:
THEOREM 1.1 Consider a K × N matrix W whose entries are independent
zero-mean complex (or real) random variables with variance σ2
N and fourth moments
of order O
1
N2
. As K , N → ∞ with K
N → α, the empirical distribution of
W WH converges almost surely to a nonrandom limiting distribution with density
f (x) = 1 −
1
α
+
δ(x) +
(x − a)+(b − x)+
2παx
where
a = σ2
(1 −
√
α)2
and b = σ2
(1 +
√
α)2
,
which is known as the Marchenko–Pastur law.
Interestingly, under the H0 hypothesis, the support of the eigenvalues of the
sample covariance matrix (in Figure 1.6, denoted by M̌P) is finite. The Marchenko–
Pastur law thus serves as a theoretical prediction under the assumption that the](https://image.slidesharecdn.com/1258189-250522005619-7be5ef49/85/Cognitive-Radio-Networks-Architectures-Protocols-And-Standards-Wireless-Networks-And-Mobile-Communications-1st-Edition-Yan-Zhang-44-320.jpg)
![22 ■ Cognitive Radio Networks
M̌P
a b
Figure 1.6 The Marchenko–Pastur support (H0 hypothesis).
matrix is “all noise.” Deviations from this theoretical limit in the eigenvalue
distribution should indicate nonnoisy components.
In the case in which a signal is present (H1), Y can be rewritten as
Y =
⎡
⎢
⎣
h1 σ 0
.
.
.
...
hK 0 σ
⎤
⎥
⎦
⎡
⎢
⎢
⎢
⎣
s(1) · · · s(N)
z1(1) · · · z1(N)
.
.
.
.
.
.
zK (1) · · · zK (N)
⎤
⎥
⎥
⎥
⎦
,
where s(k) and zi(k) = σni(k) are, respectively, the independent signal and noise
with unit variance at instant k and secondary receiver i. Let us denote by T the
matrix
T =
⎡
⎢
⎣
h1 σ 0
.
.
.
...
hK 0 σ
⎤
⎥
⎦ .
TTH clearly has one eigenvalue equal to λ1 =
|hi|2 + σ2 and all the rest
equal to σ2. The behavior of the eigenvalues of 1
N YYH is related to the study of the
eigenvalues of large sample covariance matrices of spiked population models [42].
Here, the SNR ξ is defined as
ξ =
|hi|2
σ2
.
The works by Baik et al. [42,43] have shown that when
K
N
1 and ξ
K
N
(1.15)](https://image.slidesharecdn.com/1258189-250522005619-7be5ef49/85/Cognitive-Radio-Networks-Architectures-Protocols-And-Standards-Wireless-Networks-And-Mobile-Communications-1st-Edition-Yan-Zhang-45-320.jpg)
![Spectrum Sensing in Cognitive Radio Networks ■ 23
M̌P
a b b΄
Figure 1.7 The Marchenko–Pastur support plus a signal component.
(which are assumptions that are clearly met when the number of samples N are
sufficiently high), the maximum eigenvalue of 1
N YYH converges almost surely to
b =
|hi|2
+ σ2
1 +
α
ξ
,
which is greater than the value of b = σ2(1 +
√
α)2 seen in the H0 case.
Therefore, whenever the distribution of the eigenvalues of the matrix 1
N YYH
departs from the Marchenko–Pastur law, as shown in Figure 1.7, the detector decides
that the signal is present. Hence, we apply this feature from a spectrum-sensing point
of view.
Considering λi as the eigenvalues of 1
N YYH and G = [a, b], the cooperative
sensing scheme works in two possible ways.
1.4.3.1 Noise Distribution Unknown, Variance Known
In this case, the decision criteria used is
decide for
H0: if λi ∈ G
H1: otherwise.
(1.16)
1.4.3.2 Both Noise Distribution and Variance Unknown
The ratio of the maximum and the minimum eigenvalues in the H0 hypothesis case
does not depend on the noise variance and thus serves well as a criteria independent
of the noise
decide for
H0: if λmax
λmin
≤ (1+
√
α)2
(1−
√
α)2
H1: otherwise.
(1.17)
It should be noted that, in this case, one still needs to take a sufficiently high
number of samples N such that the conditions in Equation 1.15 are met. In other
words, the number of samples scales quadratically with the inverse of the SNR.](https://image.slidesharecdn.com/1258189-250522005619-7be5ef49/85/Cognitive-Radio-Networks-Architectures-Protocols-And-Standards-Wireless-Networks-And-Mobile-Communications-1st-Edition-Yan-Zhang-46-320.jpg)
![30 ■ Cognitive Radio Networks
2.4.3 Capacity Analysis .............................................. 44
2.4.4 Results and Discussions........................................ 46
2.5 Conclusions and Open Issues ......................................... 50
2.6 Appendix: Derivation of (2.4) through (2.9) ......................... 51
References ................................................................... 52
Current static and rigid spectrum licensing policy has resulted in very inefficient spec-
trum utilization [1–3]. Cognitive radio (CR) [4–8] has been extensively researched
in recent years as a promising technology to improve spectrum utilization. The
ultimate goal of CR research is to establish a CR network that is either self-sufficient
in delivering a multitude of wireless services or capable of assisting existing wire-
less networks to enhance their performance. The performance of a CR network is
inevitably affected by the coexisting primary systems. Most importantly, the CR
transmissions should be carefully controlled to guarantee that the primary services
are not jeopardized. To better understand the ultimate performance limits and
potential applications of CR networks, it is crucial to study the CR network capacity
to provide theoretical insights into the CR network design.
In this chapter, we first introduce the classifications of CR networks. We then
analyze the capacities of two promising CR networks under average interference
power constraints. The first one is a central access CR network, which aims to provide
broadband access to CR devices with central base stations (BSs). The second one is
a cooperative CR network, where multiple dual-mode CR-cellular users collaborate
in the CR band to improve the access performance in the cellular band. Under a
simple power control framework, the uplink channel capacities of both CR networks
are analyzed and compared, taking into account various system-level factors such as
the densities and locations of primary/CR users and path loss in radio propagation
channels. Finally, the chapter concludes with some open research issues.
2.1 Classification of Cognitive Radio Networks
The core of a CR network is a coexistence mechanism that controls the spectrum
sharing in such a way that the operations of the primary system are not com-
promised. Based on different coexistence methods, CR networks can be classified
into noninterfering CR networks [9–14] and interference-tolerant CR networks
[14–17]. On the other hand, based on different radio access types [6], CR networks
can be classified as central access/infrastructure-based CR networks [9,12,13] and
ad hoc CR networks [18]. In what follows, we briefly explain these four types of CR
networks.
2.1.1 Noninterfering CR Networks
Noninterfering CR networks exploit the existence of underutilized spectrum, which
refers to the frequency segments that have been licensed to a particular primary](https://image.slidesharecdn.com/1258189-250522005619-7be5ef49/85/Cognitive-Radio-Networks-Architectures-Protocols-And-Standards-Wireless-Networks-And-Mobile-Communications-1st-Edition-Yan-Zhang-53-320.jpg)
![Capacity Analysis of Cognitive Radio Networks ■ 31
service, but are completely unused or partly utilized at a given location or a given
time. The unused frequency segments are also called frequency voids, spectrum
holes, or white spaces [7], while the partly used spectra are often referred to as
grey spaces. A noninterfering CR network seeks to collect these underutilized spectra
and reuse them on an opportunistic basis. With careful design, a noninterfering
CR network can coexist well with the primary system because it essentially seeks to
operate in a signal space orthogonal to the primary signals. A number of measurement
campaigns have shown that a large amount of white space exists in two frequency
bands: 400–800 MHz and 3–10 GHz. Therefore, noninterfering CR might start to
operate first in these two bands in the near future.
The concept of noninterfering CR networks has been widely accepted and
studied, for example, in [9–14], due to its two obvious advantages. First, the
“noninterfering” philosophy means that the primary networks can be well pro-
tected. Second, the implementation of a noninterfering CR is relatively simple.
Typically, a noninterfering CR is an intelligent wireless device that can dynamically
sense the radio spectrum, locate unused or underutilized spectrum segments (or
wireless channels) in a target spectrum pool, and automatically adjust its transceiver
parameters to communicate in the discovered free channels. Such a sensing-based
approach allows minimum changes to the primary system to tolerate CR networks.
The IEEE 802.22 working group is currently developing the first wireless standard
[9,10] based on the noninterfering CR networks. The aim is to construct a fixed
point-to-multipoint wireless regional area network (WRAN) utilizing white spaces
in the TV frequency band between 54 and 862 MHz.
2.1.2 Interference-Tolerant CR Networks
The interference-tolerant CR networks allow CR users to operate on frequency bands
assigned to the primary system as long as the total interference power received at the
primary receivers remains below a certain threshold [14–17]. As a new metric to assess
the interference at primary receivers, the concept of interference temperature [2] was
proposed by the Federal Communications Commission (FCC) in 2002. Similar to
the concept of noise temperature, interference temperature measures the power
and bandwidth occupied by interference. Moreover, the concept of interference
temperature limit [2] was introduced to characterize the “worst-case” interfering
scenario in a particular frequency band and at a particular geographic location. CR
transmissions in a given band are considered to be “harmful” only if they would raise
the interference temperature above this limit. Unlike traditional transmitter-centric
approaches that seek to regulate interference indirectly by controlling the emissions
of interfering transmitters, the interference temperature concept takes a receiver-
centric approach and aims to directly manage interference at primary receivers.
Recently, in 2007, the FCC has abandoned its use of “interference temperature”
due to current difficulties in implementing this concept. However, the philosophy](https://image.slidesharecdn.com/1258189-250522005619-7be5ef49/85/Cognitive-Radio-Networks-Architectures-Protocols-And-Standards-Wireless-Networks-And-Mobile-Communications-1st-Edition-Yan-Zhang-54-320.jpg)
![Daily
£ s. d.
15 toll-gates, at 3s.[3] 2 5 0
Hire of coach, per mile 2½d. 1 0 10
Mileage duty, 2d.[4] 0 6 8
Washing and oiling coaches 0 2 0
£ 4 8 6
For 4 weeks £ 106 4 0
Monthly.
8 road booking-offices £ 4 0 0
2 end booking-offices 2 0 0
Making Share bills 1 0 0
Oil and trimming lamps, say 0 10 0
Total £ 113 14 0
This makes £8 15s. to be divided per mile, which, of course, would
give a very handsome profit; but full loading could not be expected
every day, and if it was reduced to half loads, it would not be such a
very fat concern.
The cost of each horse was usually put at 17s. 6d. a week, including
blacksmith, and that, supposing a man to cover a ten-mile stage for
which eight horses would be ample if not running on Sundays, would
cost £7 a week, or £28 a month, leaving, at about half loading, say
£20 profit. But from this has to be deducted saddler, veterinary
surgeon, and wear and tear, the two latter of which depend, to a
certain extent, on circumstances over which he has not much
control, as it depends upon such things as sickness in the stables
and accidents.
[3]](https://image.slidesharecdn.com/1258189-250522005619-7be5ef49/85/Cognitive-Radio-Networks-Architectures-Protocols-And-Standards-Wireless-Networks-And-Mobile-Communications-1st-Edition-Yan-Zhang-63-320.jpg)
![It was usual for coaches to come to terms with the
pikers to pay for three horses instead of four.
[4]
There had also to be paid £5 licence duty yearly when
the plates were taken out.](https://image.slidesharecdn.com/1258189-250522005619-7be5ef49/85/Cognitive-Radio-Networks-Architectures-Protocols-And-Standards-Wireless-Networks-And-Mobile-Communications-1st-Edition-Yan-Zhang-64-320.jpg)
![[APPENDIX.]
G. P. O.
APPENDIX.
LIST OF MAIL COACHES WHICH WORKED OUT OF LONDON.
Bath, through { Hounslow, } From the
{ Maidenhead, } Spread Eagle,
{ Reading, }
Gracechurch
Street,
{ Newbury, } and
{ Hungerford, } Swan with Two
{ Marlborough, } Necks,
{ Devizes, } Lad Lane.
Birmingham, through { Aylesbury, } From the
{ Bicester, } King's Arms,](https://image.slidesharecdn.com/1258189-250522005619-7be5ef49/85/Cognitive-Radio-Networks-Architectures-Protocols-And-Standards-Wireless-Networks-And-Mobile-Communications-1st-Edition-Yan-Zhang-65-320.jpg)
Cognitive Radio Networks Architectures Protocols And Standards Wireless Networks And Mobile Communications 1st Edition Yan Zhang
- 1. Cognitive Radio Networks Architectures Protocols And Standards Wireless Networks And Mobile Communications 1st Edition Yan Zhang download https://ebookbell.com/product/cognitive-radio-networks- architectures-protocols-and-standards-wireless-networks-and- mobile-communications-1st-edition-yan-zhang-2516378 Explore and download more ebooks at ebookbell.com
- 2. Here are some recommended products that we believe you will be interested in. You can click the link to download. Cognitive Radio Networks From Theory To Practice 1st Edition Ahmed Khattab https://ebookbell.com/product/cognitive-radio-networks-from-theory-to- practice-1st-edition-ahmed-khattab-4230514 Cognitive Radio Networks Medium Access Control For Coexistence Of Wireless Systems 1st Edition Kaigui Bian https://ebookbell.com/product/cognitive-radio-networks-medium-access- control-for-coexistence-of-wireless-systems-1st-edition-kaigui- bian-4931202 Cognitive Radio Networks Dynamic Resource Allocation Schemes 1st Edition Shaowei Wang Auth https://ebookbell.com/product/cognitive-radio-networks-dynamic- resource-allocation-schemes-1st-edition-shaowei-wang-auth-4931920 Cognitive Radio Networks Efficient Resource Allocation In Cooperative Sensing Cellular Communications Highspeed Vehicles And Smart Grid Tao Jiang https://ebookbell.com/product/cognitive-radio-networks-efficient- resource-allocation-in-cooperative-sensing-cellular-communications- highspeed-vehicles-and-smart-grid-tao-jiang-5035112
- 3. Cognitive Radio Networks Kwangcheng Chen Ramjee Prasad https://ebookbell.com/product/cognitive-radio-networks-kwangcheng- chen-ramjee-prasad-1185266 Cooperative Cognitive Radio Networks The Complete Spectrum Cycle Mohamed Ibnkahla https://ebookbell.com/product/cooperative-cognitive-radio-networks- the-complete-spectrum-cycle-mohamed-ibnkahla-4959992 Cooperative Cognitive Radio Networks The Complete Spectrum Cycle Mohamed Ibnkahla https://ebookbell.com/product/cooperative-cognitive-radio-networks- the-complete-spectrum-cycle-mohamed-ibnkahla-4960190 Social Cognitive Radio Networks 1st Edition Xu Chen Jianwei Huang Auth https://ebookbell.com/product/social-cognitive-radio-networks-1st- edition-xu-chen-jianwei-huang-auth-4974098 Spectrum Sharing In Cognitive Radio Networks Towards Highly Connected Environments 1st Edition Prabhat Thakur https://ebookbell.com/product/spectrum-sharing-in-cognitive-radio- networks-towards-highly-connected-environments-1st-edition-prabhat- thakur-33351940
- 6. COGNITIVE RADIO NETWORKS Architectures, Protocols, and Standards
- 7. AUERBACH PUBLICATIONS www.auerbach-publications.com To Order Call: 1-800-272-7737 • Fax: 1-800-374-3401 E-mail: orders@crcpress.com Broadband Mobile Multimedia: Techniques and Applications Yan Zhang, Shiwen Mao, Laurence T. Yang, and Thomas M. Chen ISBN: 978-1-4200-5184-1 Cognitive Radio Networks: Architectures, Protocols, and Standards Yan Zhang, Jun Zheng, and Hsiao-Hwa Chen, ISBN: 978-1-4200-7775-9 Cooperative Wireless Communications Yan Zhang, Hsiao-Hwa Chen, and Mohsen Guizani ISBN: 978-1-4200-6469-8 Distributed Antenna Systems: Open Architecture for Future Wireless Communications Honglin Hu, Yan Zhang, and Jijun Luo ISBN: 978-1-4200-4288-7 The Internet of Things: From RFID to the Next-Generation Pervasive Networked Systems Lu Yan, Yan Zhang, Laurence T. Yang, and Huansheng Ning ISBN: 978-1-4200-5281-7 Millimeter Wave Technology in Wireless PAN, LAN and MAN Shao-Qiu Xiao, Ming-Tuo Zhou, and Yan Zhang ISBN: 978-0-8493-8227-7 Mobile WiMAX: Toward Broadband Wireless Metropolitan Area Networks Yan Zhang and Hsiao-Hwa Chen ISBN: 978-0-8493-2624-0 Orthogonal Frequency Division Multiple Access Fundamentals and Applications Tao Jiang, Lingyang Song, and Yan Zhang ISBN: 978-1-4200-8824-3 Resource, Mobility, and Security Management in Wireless Networks and Mobile Communications Yan Zhang, Honglin Hu, and Masayuki Fujise ISBN: 978-0-8493-8036-5 RFID and Sensor Networks: Architectures, Protocols, Security and Integrations Yan Zhang, Laurence T. Yang, and JimIng Chen ISBN: 978-1-4200-7777-3 Security in RFID and Sensor Networks Yan Zhang and Paris Kitsos ISBN: 978-1-4200-6839-9 Security in Wireless Mesh Networks Yan Zhang, Jun Zheng, and Honglin Hu ISBN: 978-0-8493-8250-5 Unlicensed Mobile Access Technology: Protocols, Architectures, Security, Standards, and Applications Yan Zhang, Laurence T. Yang, and Jianhua Ma ISBN: 978-1-4200-5537-5 WiMAX Network Planning and Optimization Yan Zhang ISBN: 978-1-4200-6662-3 Wireless Ad Hoc Networking: Personal-Area, Local-Area, and the Sensory-Area Networks Shih-Lin Wu, Yu-Chee Tseng, and Hsin-Chu ISBN: 978-0-8493-9254-2 Wireless Mesh Networking: Architectures, Protocols, and Standards Yan Zhang, Jijun Luo, and Honglin Hu ISBN: 978-0-8493-7399-2 Wireless Quality-of-Service: Techniques, Standards, and Applications Maode Ma, Mieso K. Denko, and Yan Zhang ISBN: 978-1-4200-5130-8 Dr. Yan Zhang, Series Editor Simula Research Laboratory, Norway E-mail: yanzhang@ieee.org WIRELESS NETWORKS AND MOBILE COMMUNICATIONS
- 8. COGNITIVE RADIO NETWORKS Architectures, Protocols, and Standards Edited by Yan Zhang s Jun Zheng s Hsiao-Hwa Chen
- 9. CRC Press Taylor & Francis Group 6000 Broken Sound Parkway NW, Suite 300 Boca Raton, FL 33487-2742 © 2010 by Taylor and Francis Group, LLC CRC Press is an imprint of Taylor & Francis Group, an Informa business No claim to original U.S. Government works Printed in the United States of America on acid-free paper 10 9 8 7 6 5 4 3 2 1 International Standard Book Number-13: 978-1-4200-7776-6 (Ebook-PDF) This book contains information obtained from authentic and highly regarded sources. Reasonable efforts have been made to publish reliable data and information, but the author and publisher cannot assume responsibility for the validity of all materials or the consequences of their use. The authors and publishers have attempted to trace the copyright holders of all material reproduced in this publication and apologize to copyright holders if permission to publish in this form has not been obtained. If any copyright material has not been acknowledged please write and let us know so we may rectify in any future reprint. Except as permitted under U.S. Copyright Law, no part of this book may be reprinted, reproduced, transmit- ted, or utilized in any form by any electronic, mechanical, or other means, now known or hereafter invented, including photocopying, microfilming, and recording, or in any information storage or retrieval system, without written permission from the publishers. For permission to photocopy or use material electronically from this work, please access www.copyright. com (http://www.copyright.com/) or contact the Copyright Clearance Center, Inc. (CCC), 222 Rosewood Drive, Danvers, MA 01923, 978-750-8400. CCC is a not-for-profit organization that provides licenses and registration for a variety of users. For organizations that have been granted a photocopy license by the CCC, a separate system of payment has been arranged. Trademark Notice: Product or corporate names may be trademarks or registered trademarks, and are used only for identification and explanation without intent to infringe. Visit the Taylor & Francis Web site at http://www.taylorandfrancis.com and the CRC Press Web site at http://www.crcpress.com
- 10. Contents Preface ........................................................................ vii Editors ........................................................................ ix Contributors .................................................................. xv PART I PHYSICAL LAYER ISSUES 1 Spectrum Sensing in Cognitive Radio Networks......................... 3 LEONARDO S. CARDOSO, MÉROUANE DEBBAH, SAMSON LASAULCE, MARI KOBAYASHI, AND JACQUES PALICOT 2 Capacity Analysis of Cognitive Radio Networks ......................... 29 XUEMIN HONG, CHENG-XIANG WANG, JOHN THOMPSON, AND HSIAO-HWA CHEN 3 Power Control for Cognitive Radio Ad Hoc Networks .................. 57 LIJUN QIAN, XIANGFANG LI, JOHN ATTIA, AND DEEPAK KATARIA PART II PROTOCOLS AND ECONOMIC APPROACHES 4 Medium Access Control in Cognitive Radio Networks .................. 89 JIE XIANG AND YAN ZHANG 5 Cross-Layer Optimization in Cognitive Radio Networks ............... 121 CHRISTIAN DOERR, DIRK GRUNWALD, AND DOUGLAS C. SICKER 6 Security in Cognitive Radio Networks.................................... 161 JACK L. BURBANK 7 Distributed Coordination in Cognitive Radio Networks................ 183 CHRISTIAN DOERR, DOUGLAS C. SICKER, AND DIRK GRUNWALD 8 Quality-of-Service in Cognitive WLAN over Fiber ...................... 221 HAOMING LI, QIXIANG PANG, AND VICTOR C. M. LEUNG v
- 11. vi ■ Contents 9 Game Theory for Dynamic Spectrum Access............................. 259 SAMIR MEDINA PERLAZA, SAMSON LASAULCE, MÉROUANE DEBBAH, AND JEAN-MARIE CHAUFRAY 10 Game Theory for Spectrum Sharing ...................................... 291 JIANWEI HUANG AND ZHU HAN 11 Pricing for Security and QoS in Cognitive Radio Networks ............ 319 S. SENGUPTA, S. ANAND, AND R. CHANDRAMOULI PART III APPLICATIONS AND SYSTEMS 12 Cognitive Radio for Pervasive Healthcare ................................ 353 PHOND PHUNCHONGHARN, EKRAM HOSSAIN, AND SERGIO CAMORLINGA 13 Network Selection in Cognitive Radio Networks ........................ 393 YONG BAI, YIFAN YU, AND LAN CHEN 14 Cognitive Radio Networks: An Assessment Framework ................. 423 MIKHAIL SMIRNOV, JENS TIEMANN, AND KLAUS NOLTE Index ............................................................................ 455
- 12. Preface Spectrum is a scarce and precious resource in wireless communication systems and networks. Currently, wireless networks are regulated by a fixed spectrum assign- ment policy. This strategy partitions the spectrum into a large number of different ranges. Each piece is specified for a particular system. This leads to the undesirable situation that some systems may use only the allocated spectrum to a very limited extent while others have very serious spectrum insufficiency problems. In addition, future-generation broadband wireless networking promises to provide broadband multimedia services under heterogeneous networks coexistence. These challenges and requirements make the problem of scarce spectra even worse, and motivate new technologies to efficiently use spectra and combat the vulnerability of wireless channels. Cognitive radio is believed to be a high-potential technology to address these issues. It refers to the potentiality that systems are aware of context and are capable of reconfiguring themselves based on the surrounding environments and their own properties with respect to spectrum, traffic load, congestion situation, network topology, and wireless channel propagation. This capability is particularly applicable to resolve heterogeneity, robustness, and openness. However, cognitive wireless networks are still in the very early stages of research and development. There are a number of technical, economical, and regulatory challenges to be addressed. In addition, there are unique complexities in aspects of spectrum sensing, spectrum management, spectrum sharing, and spectrum mobility. This book systematically introduces and explains cognitive radio wireless net- works. It provides a comprehensive technical guide covering introductory concepts, fundamental techniques, recent advances, and open issues in cognitive radio commu- nications and networks. It also contains illustrative figures and allows for complete cross-referencing. This book is organized into three parts: ■ Part I: Physical Layer Issues ■ Part II: Protocols and Economic Approaches ■ Part III: Applications and Systems vii
- 13. viii ■ Preface Part I introduces the issues and solutions in the physical layer, including sensing, capacity, and power control. Part II introduces the issues and solutions in the protocol layers. This part also contributes to the applications of economic approaches in cognitive radio networks. Part III explores applications and practical cognitive radio systems. This book has the following salient features: ■ It serves as a comprehensive and essential reference on cognitive radio. ■ It covers basics, a broad range of topics, and future development directions. ■ It introduces architectures, protocols, security, and applications. ■ It assists professionals, engineers, students, and researchers This book can serve as an essential reference for students, educators, research strate- gists, scientists, researchers, and engineers in the field of wireless communications and networking. In particular, it will have an instant appeal to students, researchers, developers, and consultants in developing future-generation wireless systems and networks. The content in this book will enable readers to understand the neces- sary background, concepts, and principles in the framework of cognitive wireless systems. It will also provide readers with a comprehensive technical guidance on cognitive radio, cognitive wireless networks, and dynamic spectrum access. The issues covered include spectrum sensing, medium access control (MAC), cooper- ation schemes, resource management, mobility, game theoretical approach, and healthcare application. We would like to acknowledge the time and effort invested by the contributors for their excellent work. All of them were extremely professional and cooperative. Special thanks go to Richard O’Hanley, Stephanie Morkert, and Joette Lynch of Taylor & Francis Group for their patience, support, and professionalism from the beginning until the final stage. We are very grateful to Sathyanarayanamoorthy Sridharan at SPi for his great efforts during the production process. Last but not least, a special thank you to our families and friends for their constant encouragement, patience, and understanding throughout this project. Yan Zhang Simula Research Laboratory, Norway Jun Zheng Southeast University, China Hsiao-Hwa Chen National Cheng Kung University, Taiwan
- 14. Editors Yan Zhang received a BS in communication engineering from the Nanjing University of Post and Telecommunications, China; an MS in electrical engineering from the Beijing University of Aeronautics and Astronautics, China; and a PhD from the School of Electrical & Electronics Engineering, Nanyang Technological University, Singapore. He is an associate editor or editorial board member of Wiley’s International Journal of Communication Systems (IJCS); the International Journal of Communica- tion Networks and Distributed Systems (IJCNDS); Springer’s International Journal of Ambient Intelligence and Humanized Computing (JAIHC); the International Journal of Adaptive, Resilient and Autonomic Systems (IJARAS); Wiley’s Wireless Commu- nications and Mobile Computing (WCMC); Wiley’s Security and Communication Networks; the International Journal of Network Security; the International Jour- nal of Ubiquitous Computing; Transactions on Internet and Information Systems (TIIS); the International Journal of Autonomous and Adaptive Communications Sys- tems (IJAACS); the International Journal of Ultra Wideband Communications and Systems (IJUWBCS); and the International Journal of Smart Home (IJSH). He is currently serving as an editor for the book series Wireless Networks and Mobile Communications (Auerbach Publications, CRC Press, Taylor & Francis Group). He serves as a guest coeditor for Wiley’s Wireless Communications and Mobile Computing (WCMC) special issue for best papers in the conference IWCMC 2009; ACM/Springer’s Multimedia Systems Journal special issue on “wireless mul- timedia transmission technology and application”; Springer’s Journal of Wireless Personal Communications special issue on “cognitive radio networks and com- munications”; Inderscience’s International Journal of Autonomous and Adaptive Communications Systems (IJAACS) special issue on “ubiquitous/pervasive services and applications”; EURASIP’s Journal on Wireless Communications and Networking (JWCN) special issue on “broadband wireless access”; IEEE Intelligent Systems special issue on “context-aware middleware and intelligent agents for smart environments”; Wiley’s Security and Communication Networks special issue on “secure multimedia communication”; Elsevier’s Computer Communications special issue on “adaptive multicarrier communications and networks”; Inderscience’s International Journal of ix
- 15. x ■ Editors Autonomous and Adaptive Communications Systems (IJAACS) special issue on “cogni- tive radio systems”; the Journal of Universal Computer Science (JUCS) special issue on “multimedia security in communication”; Springer’s Journal of Cluster Computing special issue on “algorithm and distributed computing in wireless sensor networks”; EURASIP’s Journal on Wireless Communications and Networking (JWCN) special issue on “OFDMA architectures, protocols, and applications”; and Springer’s Jour- nal of Wireless Personal Communications special issue on “security and multimodality in pervasive environments.” He is also serving as a coeditor for several books, including Resource, Mobility and Security Management in Wireless Networks and Mobile Communications; Wireless Mesh Networking: Architectures, Protocols and Standards; Millimeter-Wave Technology in Wireless PAN, LAN and MAN; Distributed Antenna Systems: Open Architecture for Future Wireless Communications; Security in Wireless Mesh Networks; Mobile WiMAX: Toward Broadband Wireless Metropolitan Area Networks; Wireless Quality- of-Service: Techniques, Standards and Applications; Broadband Mobile Multimedia: Techniques and Applications; Internet of Things: From RFID to the Next-Generation Pervasive Networked Systems; Unlicensed Mobile Access Technology: Protocols, Archi- tectures, Security, Standards and Applications; Cooperative Wireless Communications; WiMAX Network Planning and Optimization; RFID Security: Techniques, Proto- cols and System-on-Chip Design; Autonomic Computing and Networking; Security in RFID and Sensor Networks; Handbook of Research on Wireless Security; Handbook of Research on Secure Multimedia Distribution; RFID and Sensor Networks; Cog- nitive Radio Networks; Wireless Technologies for Intelligent Transportation Systems; Vehicular Networks: Techniques, Standards and Applications; Orthogonal Frequency Division Multiple Access (OFDMA); Game Theory for Wireless Communications and Networking; and Delay Tolerant Networks: Protocols and Applications. He serves or has served as industrial liaison cochair for UIC 2010, program cochair for WCNIS 2010, symposium vice chair for CMC 2010, program track chair for BodyNets 2010, program chair for IWCMC 2010, program cochair for WICON 2010, program vice chair for CloudCom 2009, publicity cochair for IEEE MASS 2009, publicity cochair for IEEE NSS 2009, publication chair for PSATS 2009, symposium cochair for ChinaCom 2009, program cochair for BROADNETS 2009, program cochair for IWCMC 2009, workshop cochair for ADHOCNETS 2009, general cochair for COGCOM 2009, program cochair for UC-Sec 2009, journal liasion chair for IEEE BWA 2009, track cochair for ITNG 2009, publicity cochair for SMPE 2009, publicity cochair for COMSWARE 2009, publicity cochair for ISA 2009, general cochair for WAMSNet 2008, publicity cochair for TrustCom 2008, general cochair for COGCOM 2008, workshop cochair for IEEE APSCC 2008, general cochair for WITS-08, program cochair for PCAC 2008, general cochair for CONET 2008, workshop chair for SecTech 2008, workshop chair for SEA 2008, workshop co-organizer for MUSIC’08, workshop co-organizer for 4G-WiMAX 2008, publicity cochair for SMPE-08, international journals coordinating cochair for FGCN-08, publicity cochair for ICCCAS 2008, workshop chair for ISA 2008,
- 16. Editors ■ xi symposium cochair for ChinaCom 2008, industrial cochair for MobiHoc 2008, program cochair for UIC-08, general cochair for CoNET 2007, general cochair for WAMSNet 2007, workshop cochair for FGCN 2007, program vice cochair for IEEE ISM 2007, publicity cochair for UIC-07, publication chair for IEEE ISWCS 2007, program cochair for IEEE PCAC’07, special track cochair for Mobility and Resource Management in Wireless/Mobile Networks in ITNG 2007, special session co-organizer for Wireless Mesh Networks in PDCS 2006, a member of the Tech- nical Program Committee for numerous international conferences, including ICC, GLOBECOM, WCNC, PIMRC, VTC, CCNC, AINA, ISWCS, etc. He received the Best Paper Award in the IEEE 21st International Conference on Advanced Information Networking and Applications (AINA-07). Since August 2006, he has been working with Simula Research Laboratory, Lysaker, Norway (http://www.simula.no/). His research interests include resource, mobility, spectrum, data, energy, and security management in wireless networks and mobile computing. He is a member of IEEE and IEEE ComSoc. Jun Zheng is a full professor with the National Mobile Communications Research Laboratory at Southeast University, Nanjing, China. He received a PhD in electrical and electronic engineering from the University of Hong Kong, China. Before joining Southeast University, he was with the School of Information Technology and Engineering of the University of Ottawa, Canada. Dr. Zheng serves as a technical editor of IEEE Communications Magazine and IEEE Communications Surveys & Tutorials. He is also the founding editor in chief of ICST Transactions on Mobile Communications and Applications, and an editorial board member of several other refereed journals, including Wiley’s Wireless Com- munications and Mobile Computing, Wiley’s Security and Communication Networks, Inderscience’s International Journal of Communication Networks and Distributed Systems, and Inderscience’s International Journal of Autonomous and Adaptive Com- munications Systems. He has coedited eight special issues for different refereed journals and magazines, including IEEE Journal on Selected Areas in Communications, IEEE Network, Wiley’s Wireless Communications and Mobile Computing, Wiley’s Inter- national Journal of Communication Systems, and Springer’s Mobile Networks and Applications, all as lead guest editor. Dr. Zheng has served as general chair of AdHoctNets’09 and AccessNets’07, TPC cochair of AdHocNets’10 and AccessNets’08, and symposium cochair of IEEE GLOBECOM’08, ICC’09, GLOBECOM’10, and ICC’11. He is also serving on the steering committees of AdHocNets and AccessNets, and has served on the technical program committees of a number of international conferences and symposia, including IEEE ICC and GLOBECOM. Dr. Zheng has conducted extensive research in the field of communication net- works. The scope of his research includes design and analysis of network architecture and protocols for efficient and reliable communications, and their applications to different types of communication networks, covering wireless networks and wired
- 17. xii ■ Editors networks. His current research interests are focused on mobile communications and wireless ad hoc networks. He has coauthored books published by Wiley–IEEE Press, and has published a number of technical papers in refereed journals and magazines as well as in peer-reviewed conference proceedings. He is a senior member of the IEEE. Hsiao-Hwa Chen is currently a full professor in the Department of Engineering Science, National Cheng Kung University, Tainan, Taiwan. He received a BSc and MSc with the highest honor from Zhejiang University, Hangzhou, China, and a PhD from the University of Oulu, Finland, in 1982, 1985, and 1990, respectively, all in electrical engineering. He worked with the Academy of Finland as a research associate from 1991 to 1993, and with the National University of Singapore as a lecturer and then as a senior lecturer from 1992 to 1997. He joined the Department of Electrical Engineering, National Chung Hsing University, Taichung, Taiwan, as an associate professor in 1997 and was promoted to a full professor in 2000. In 2001, he joined National Sun Yat-Sen University, Kaohsiung, Taiwan, as the founding chair of the Institute of Communications Engineering of the university. Under his strong leadership, the institute was ranked second in the country in terms of SCI journal publications and National Science Council funding per faculty member in 2004. In particular, National Sun Yat-Sen University was ranked first in the world in terms of the number of SCI journal publications in wireless LAN research papers during 2004 to mid-2005, according to a research report released by The Office of Naval Research, United States. He was a visiting professor to the Department of Electrical Engineering, University of Kaiserslautern, Germany, in 1999; the Institute of Applied Physics, Tsukuba University, Japan, in 2000; the Institute of Experimental Mathematics, University of Essen, Germany, in 2002 (under DFG Fellowship); the Chinese University of Hong Kong in 2004; and the City University of Hong Kong in 2007. His current research interests include wireless networking, MIMO systems, infor- mation security, and Beyond 3G wireless communications. He is the inventor of next-generation CDMA technologies. He is also a recipient of numerous research and teaching awards from the National Science Council, the Ministry of Education, and other professional groups in Taiwan. He has authored or coauthored over 200 technical papers in major international journals and conferences, and five books and several book chapters in the area of communications, including Next Generation Wireless Systems and Networks and The Next Generation CDMA Technologies, both of which were published by Wiley in 2005 and 2007, respectively. He has been an active volunteer for IEEE for various technical activities for over 15 years. Currently, he is serving as the chair of IEEE Communications Society Radio Communications Committee, and the vice chair of IEEE Communications Society Communications & Information Security Technical Committee. He served or is serving as symposium chair/cochair of many major IEEE conferences, includ- ing IEEE VTC 2003 Fall, IEEE ICC 2004, IEEE Globecom 2004, IEEE ICC
- 18. Editors ■ xiii 2005, IEEE Globecom 2005, IEEE ICC 2006, IEEE Globecom 2006, IEEE ICC 2007, IEEE WCNC 2007, etc. He served or is serving as an editorial board mem- ber and/or guest editor of IEEE Communications Letters, IEEE Communications Magazine, IEEE Wireless Communications Magazine, IEEE JSAC, IEEE Network Magazine, IEEE Transactions on Wireless Communications, and IEEE Vehicular Technology Magazine. He is the editor in chief of Wiley’s Security and Commu- nication Networks journal (www.interscience.wiley.com/journal/security), and the special issue editor in chief of Hindawi Journal of Computer Systems, Networks, and Communications (http://www.hindawi.com/journals/jcsnc/). He is also serv- ing as the chief editor (Asia and Pacific) for Wiley’s Wireless Communications and Mobile Computing (WCMC) journal and International Journal of Communication Systems. His original work in CDMA wireless networks, digital communications, and radar systems has resulted in five U.S. patents, two Finnish patents, three Taiwanese patents, and two Chinese patents, some of which have been licensed to industry for commercial applications. He is an adjunct professor of Zhejiang University, China, and Shanghai Jiao Tong University, China. Professor Chen is the recipient of the Best Paper Award in IEEE WCNC 2008 and he is also a fellow of IEEE and IET.
- 20. Contributors S. Anand Department of Electrical and Computer Engineering Stevens Institute of Technology Hoboken, New Jersey John Attia Department of Electrical and Computer Engineering Prairie View A&M University Prairie View, Texas Yong Bai DOCOMO Beijing Communications Labs Beijing, China Jack L. Burbank Applied Physics Laboratory Johns Hopkins University Baltimore, Maryland Sergio Camorlinga Departments of Radiology and Computer Science University of Manitoba and TRLabs Winnipeg, Manitoba, Canada Leonardo S. Cardoso SUPELEC Gif-sur-Yvette, France R. Chandramouli Department of Electrical and Computer Engineering Stevens Institute of Technology Hoboken, New Jersey Jean-Marie Chaufray Orange Labs France Telecom R&D Paris, France Hsiao-Hwa Chen Department of Engineering Science National Cheng Kung University Tainan, Taiwan Lan Chen DOCOMO Beijing Communications Labs Beijing, China Mérouane Debbah SUPELEC Gif-sur-Yvette, France xv
- 21. xvi ■ Contributors Christian Doerr Department of Computer Science University of Colorado Boulder, Colorado and Department of Telecommunications Technische Universiteit Delft Delft, the Netherlands Dirk Grunwald Department of Computer Science University of Colorado Boulder, Colorado Zhu Han Department of Electrical and Computer Engineering University of Houston Houston, Texas Xuemin Hong Joint Research Institute for Signal and Image Processing School of Engineering and Physical Sciences Heriot-Watt University Edinburgh, United Kingdom Ekram Hossain Department of Electrical and Computer Engineering University of Manitoba and TRLabs Winnipeg, Manitoba, Canada Jianwei Huang Department of Information Engineering The Chinese University of Hong Kong Hong Kong, People’s Republic of China Deepak Kataria LSI Corporation Allentown, Pennsylvania Mari Kobayashi SUPELEC Gif-sur-Yvette, France Samson Lasaulce Laboratoire des Signaux et Systèmes Centre National de la Recherche Scientifique SUPELEC Gif-sur-Yvette, France Victor C. M. Leung Department of Electrical and Computer Engineering The University of British Columbia Vancouver, British Columbia, Canada Haoming Li Department of Electrical and Computer Engineering The University of British Columbia Vancouver, British Columbia, Canada Xiangfang Li Department of Electrical and Computer Engineering Texas A&M University College Station, Texas
- 22. Contributors ■ xvii Klaus Nolte Alcatel-Lucent Deutschland AG Bell Labs Stuttgart, Germany Jacques Palicot Signal, Communication et Electronique Embarquée SUPELEC Rennes, France Qixiang Pang General Dynamics Canada Calgary, Alberta, Canada Samir Medina Perlaza Orange Labs France Telecom R&D Paris, France Phond Phunchongharn Department of Electrical and Computer Engineering University of Manitoba and TRLabs Winnipeg, Manitoba, Canada Lijun Qian Department of Electrical and Computer Engineering Prairie View A&M University Prairie View, Texas S. Sengupta Department of Mathematics and Computer Science City University of New York New York, New York Douglas C. Sicker Department of Computer Science University of Colorado Boulder, Colorado Mikhail Smirnov Fraunhofer Institute for Open Communication Systems Berlin, Germany John Thompson Institute for Digital Communications Joint Research Institute for Signal and Image Processing School of Engineering and Electronics The University of Edinburgh Edinburgh, United Kingdom Jens Tiemann Fraunhofer Institute for Open Communication Systems Berlin, Germany Cheng-Xiang Wang Joint Research Institute for Signal and Image Processing School of Engineering and Physical Sciences Heriot-Watt University Edinburgh, United Kingdom Jie Xiang Simula Research Laboratory Lysaker, Norway Yifan Yu DOCOMO Beijing Communications Labs Beijing, China Yan Zhang Simula Research Laboratory Lysaker, Norway
- 26. Chapter 1 Spectrum Sensing in Cognitive Radio Networks Leonardo S. Cardoso, Mérouane Debbah, Samson Lasaulce, Mari Kobayashi, and Jacques Palicot Contents 1.1 Introduction ........................................................... 4 1.1.1 Interference Management and Spectrum Sensing ............. 5 1.1.1.1 Receiver-Centric Interference Management ......... 5 1.1.1.2 Transmitter-Centric Interference Management ..... 5 1.1.2 Characteristics of Spectrum Sensing........................... 6 1.2 Problem Formulation ................................................. 6 1.2.1 The General Spectrum-Sensing Problem...................... 6 1.2.2 Spectrum Sensing from the Cognitive Radio Network Perspective ..................................................... 8 1.2.2.1 No Prior Knowledge on the Signal Structure........ 9 1.2.2.2 Sensing Time......................................... 9 1.2.2.3 Fading Channels ..................................... 9 3
- 27. 4 ■ Cognitive Radio Networks 1.3 Noncooperative Sensing Techniques ................................. 9 1.3.1 Energy Detector ............................................... 10 1.3.1.1 Characterization of Energy Detector in AWGN Channels ............................................. 11 1.3.1.2 Characterization of Energy Detector in Fading Channels ............................................. 12 1.3.2 Matched Filter Detector ....................................... 12 1.3.2.1 Characterization of the Matched Filter .............. 14 1.3.3 Cyclostationary Feature Detection ............................ 14 1.4 Cooperative Sensing Techniques...................................... 15 1.4.1 Voting-Based Sensing.......................................... 17 1.4.2 Correlator-Based Sensing ...................................... 19 1.4.3 Eigenvalue-Based Sensing ..................................... 20 1.4.3.1 Noise Distribution Unknown, Variance Known .... 23 1.4.3.2 Both Noise Distribution and Variance Unknown... 23 1.5 Conclusions and Open Issues ......................................... 24 References ................................................................... 25 Today, the creation of new radio access technologies is limited by the shortage of the available radio spectrum. These new technologies are becoming evermore bandwidth demanding due to their higher rate requirements. Cognitive radio networks and spectrum-sensing techniques are a natural way to allow these new technologies to be deployed. In this chapter, we discuss spectrum sensing for cognitive radio networks. We begin by introducing the subject in Section 1.1, providing a brief background followed by a discussion of spectrum-sensing motivations and characteristics. Then we move on to the spectrum-sensing problem itself in Section 1.2, where we explain the issues that are inherent to spectrum sensing. In Section 1.3, we explore the classical noncooperative spectrum-sensing techniques that form the basis for the more elaborate, cooperative techniques presented in Section 1.4. Finally, we close this chapter with some conclusions and open issues. 1.1 Introduction One of the most prominent features of cognitive radio networks will be the ability to switch between radio access technologies, transmitting in different portions of the radio spectrum as unused frequency band slots arise [1–3]. This dynamic spectrum access is one of the fundamental requirements for transmitters to adapt to varying channel qualities, network congestion, interference, and service requirements. Cog- nitive radio networks (from now on called secondary networks) will also need to coexist with legacy ones (hereafter called primary networks), which have the right to their spectrum slice and thus cannot accept interference.
- 28. Spectrum Sensing in Cognitive Radio Networks ■ 5 Based on these facts, underutilization of the current spectrum and the need to increase the network capacity is pushing research toward new means of exploiting the wireless medium. In this direction, the Federal Communications Commission (FCC) Spectrum Policy Task Force has published a report [4] in 2002, in which it thoroughly investigates the underutilization of the radio spectrum. While the FCC is in charge of determining the spectrum usage and its policies, the Whitespace Coalition is studying ways to exploit the spectrum vacancies in the television band. Cognitive radio networks are envisioned to be able to opportunistically exploit those spectrum “leftovers,” by means of knowledge of the environment and cognition capability, to adapt to their radio parameters accordingly. Spectrum sensing is the technique that will enable cognitive radio networks to achieve this goal. 1.1.1 Interference Management and Spectrum Sensing To share the spectrum with legacy systems, cognitive radio networks will have to respect some set of policies defined by regulatory agencies [2,3]. These policies are based on the central idea where there are primary systems that have the right to the spectrum and secondary systems that are allowed to use the spectrum so long as they do not disturb the communications of the primary systems. Roughly speaking, these policies deal with controlling the amount of interference that the secondary systems can incur to primary ones. Thus, the problem is one of interference management [2,3]. We can address this problem from two different points of view: receiver centric or transmitter centric. 1.1.1.1 Receiver-Centric Interference Management In the receiver-centric approach [2,3], an interference limit at the receiver is calcu- lated and used to determine the restriction on the power of the transmitters around it. This interference limit, called the interference temperature, is chosen to be the worst interference level that can be accepted without disturbing the receiver operation beyond its operating point. Although very interesting, this approach requires knowl- edge of the interference limits of all receivers in a primary system. Such knowledge depends on many variables, including individual locations, fading situations, mod- ulations, coding schemes, and services. Receiver-centric interference-management techniques are not addressed in this chapter as they have been recently ruled out by the IEEE SCC41 cognitive radio network standard. 1.1.1.2 Transmitter-Centric Interference Management In the transmitter-centric approach, the focus is shifted to the source of interference [2,3]. The transmitter does not know the interference temperature, but by means of sensing, it tries to detect free bandwidth. The sensing procedure allows the transmitter to classify the channel status to decide whether it can transmit and with
- 29. 6 ■ Cognitive Radio Networks how much power. In actual systems, however, as the transmitter does not know the location of the receivers or their channel conditions, it is not able to infer how much interference these receivers can tolerate. Thus, spectrum sensing solves the problem for worst-case scenario, assuming strong interference channels, so that the secondary system transmits only when it senses an empty medium. 1.1.2 Characteristics of Spectrum Sensing There are several techniques available for spectrum sensing, each with its own set of advantages and disadvantages that depend on the specific scenario. Some works in the literature [5–7] consider spectrum sensing as a method for distinguishing between two or more different types of signals or technologies in operation. Because this is not a question of detection (determining whether a given frequency band is being used), these types of signal identification issues [8] are not addressed in this chapter. Rather we focus on their detection. Ultimately, a spectrum-sensing device must be able to give a general picture of the medium over the entire radio spectrum. This allows the cognitive radio network to analyze all degrees of freedom (time, frequency, and space) to predict the spectrum usage. Wideband spectrum-sensing works are also available in the literature [9–12]; however, an equipment able to perform wideband sensing all at once is prohibitively difficult to build with today’s technology. Feasible spectrum-sensing devices can quickly sweep the radio spectrum, analyzing one narrowband segment at a time. This chapter focuses on narrowband-sensing techniques. In this section, we have emphasized the importance of the spectrum-sensing technique for cognitive radio networks. In the next section, we aim at understanding the underlying characteristics of the spectrum-sensing problem, which will enable us to develop the approaches presented further in this chapter. 1.2 Problem Formulation 1.2.1 The General Spectrum-Sensing Problem Spectrum sensing is based on a well-known technique called signal detection. In a nutshell, signal detection can be described as a method for identifying the presence of a signal in a noisy environment. Signal detection has been thoroughly studied for radar purposes since the 1950s [13]. Analytically, signal detection can be reduced to a simple identification problem, formalized as a hypothesis test [14–16]: y(k) = n(k): H0 s(k) + n(k): H1 , (1.1) where y(k) is the sample to be analyzed at each instant k n(k) is the noise (not necessarily white Gaussian noise) of variance σ2
- 30. Spectrum Sensing in Cognitive Radio Networks ■ 7 H0 H1 H0 H1 P(H0|H0) P(H0|H1) P(H1|H0) P(H1|H1) Figure 1.1 Hypothesis test and possible outcomes with their corresponding probabilities. s(k) is the signal the network wants to detect H0 and H1 are the noise-only and signal-plus-noise hypotheses, respectively H0 and H1 are the sensed states for the absence and presence of signal, respec- tively. Then, as shown in Figure 1.1 we can define four possible cases for the detected signal: 1. Declaring H0 when H0 is true (H0|H0) 2. Declaring H1 when H1 is true (H1|H1) 3. Declaring H0 when H1 is true (H0|H1) 4. Declaring H1 when H0 is true (H1|H0) Case 2 is known as a correct detection, whereas cases 3 and 4 are known as a missed detection and a false alarm, respectively. Clearly, the aim of the signal detector is to achieve correct detection all of the time, but this can never be perfectly achieved in practice because of the statistical nature of the problem. Therefore, signal detectors are designed to operate within prescribed minimum error levels. Missed detections are the biggest issue for spectrum sensing, as it means possibly interfering with the primary system. Nevertheless, it is desirable to keep the false alarm rate as low as possible for spectrum sensing, so that the system can exploit all possible transmission opportunities. The performance of the spectrum-sensing technique is usually influenced by the probability of false alarm Pf = P(H1|H0), because this is the most influential metric. Usually, the performance is presented by receiver operation characteristic (ROC) curves, which plot the probability of detection Pd = P(H1|H1) as a function of the probability of false alarm Pf . Equation 1.1 shows that, to distinguish H0 and H1, a reliable way to differentiate signal from noise is required. This becomes very difficult in the case where the statistics of the noise are not well known or when the signal-to-noise ratio (SNR) is low, in which case the signal characteristics are buried under the noise, as shown by Tandra et al. in [17]. In fact, this work also shows that the lesser one knows about
- 31. 8 ■ Cognitive Radio Networks the statistics of the noise, the worse the performance of any signal detector is in the low-SNR regime. Clearly, the noise characteristics are very important for the spectrum-sensing procedure. Most works on spectrum sensing consider noise to be additive white Gaussian noise (AWGN), because many independent sources of noise are added (central limit theory). Nevertheless, in realistic scenarios, this approximation may not be appropriate, because receivers modify the noise through processes such as filters, amplifier nonlinearities, and automatic gain controls [18,19]. Poor performance in a low-SNR regime means that all of the techniques available are negatively affected by poor channels. In the case of variable channel gains, Equation 1.1 is rewritten as y(k) = n(k): H0 h(k)s(k) + n(k): H1 , (1.2) where h(k) is the channel gain at each instant k. In a wireless radio network, as it is reasonable to assume that the spectrum-sensing device does not know the location of the transmitter, two options arise: ■ A low h(k) is solely due to the pathloss (distance) between the transmitter and the sensing device, meaning that the latter is out of range. ■ A low h(k) is due to shadowing or multipath, meaning that the sensing device might be within the range of the transmitter. In the latter case, a critical issue arises. Therein, the fading plays an especially negative role in the well-known “hidden node” problem [20]. In this problem, the spectrum-sensing terminal is deeply faded with respect to the transmitting node while having a good channel to the receiving node. The spectrum-sensing node then senses a free medium and initiates its transmission, which produces interference on the primary transmission. Thus, fading here introduces uncertainty regarding the estimation problem. To solve this issue, cooperative sensing has been proposed. In this approach, several sensing terminals gather their information to make a joint decision about the medium availability. Cooperative spectrum sensing is further explored in Section 1.4. 1.2.2 Spectrum Sensing from the Cognitive Radio Network Perspective In contrast to the general case, where only the signal detection aspect is considered, the problem of spectrum sensing as seen from a cognitive radio perspective has very stringent restrictions. These are mainly imposed by the policies these cognitive
- 32. Spectrum Sensing in Cognitive Radio Networks ■ 9 radio networks face to be able to operate alongside legacy networks. Some of these restrictions are summarized in Sections 1.2.2.1 through 1.2.2.3. 1.2.2.1 No Prior Knowledge on the Signal Structure There are portions of the spectrum where multiple technologies (using different protocols) share the spectrum, such as the ones operating on the instrumentation scientific and medical (ISM) unlicensed band. Cognitive radio networks must be able to deal with existing multiple technologies, as well as new ones that may eventually appear across the span of the wireless radio spectrum. These networks should be able to discover the state of the medium irrespective of the technologies in use. Of course, if the technologies are known, then this information can be exploited to improve the accuracy of the spectrum sensing, for example, through the detection of known pilot sequences within the signal [17]. 1.2.2.2 Sensing Time Due to the primary importance of the legacy system, the secondary system must be designed to free the medium as soon as it senses that a legacy network has initiated a transmission. For efficient use of the spectrum, these secondary networks must also sense available spectrum as quickly as possible, in the least possible number of received samples. In general terms, spectrum-sensing techniques work through a compromise between the number of samples and accuracy. Cooperative spectrum sensing gives the opportunity to decrease the sensing time for the same level of accuracy. 1.2.2.3 Fading Channels As discussed earlier, spectrum sensing is particularly sensitive to fading environments. Communication systems operate in diverse environments, including those prone to fading. Thus, in many situations, spectrum-sensing devices must be able to detect reliably even over heavily faded channels. Although several works have focused on sensing for the fading environment in the noncooperative setting [21,22], it is foreseen that cooperative sensing [23–31] is the best way to address this problem. Nevertheless, it creates other implications such as the distribution of metrics among the sensing terminals and the decision regarding which terminals are to be considered dependable or not. 1.3 Noncooperative Sensing Techniques In a realistic spectrum-sensing scenario, there are situations in which only one sensing terminal is available or in which no cooperation is allowed due to the lack
- 33. 10 ■ Cognitive Radio Networks of communication between sensing terminals. In this section, we explore the main single-user sensing schemes, some of which will serve as a basis for the development of the cooperative ones investigated in Section 1.4. Single-user spectrum-sensing approaches have been widely studied in the lit- erature, in part because of the relationship to signal detection. There are several classical techniques for this purpose, including the energy detector (ED) [16,21,22], the matched filter (MF) [25,32], and the cyclostationary feature detection (CFD) [6,33–36]. 1.3.1 Energy Detector The most well-known spectrum-sensing technique is the ED. It is based on the principle that, at the reception, the energy of the signal to be detected is always higher than the energy of the noise. The ED is said to be a blind signal detector because it ignores the structure of the signal. It estimates the presence of a signal by comparing the energy received with a known threshold ν [16,21,22], derived from the statistics of the noise. Let y(k) be a sequence of received samples k ∈ {1, 2, . . . , N} at the signal detector, such as that in Equation 1.1. Then, the decision rule can be stated as decide for H0, if E ν H1, if E ≥ ν , where E = E[| y(k) |2] is the estimated energy of the received signal ν is chosen to be the noise variance σ2 In practice, one does not dispose of the actual received energy power E. The ED uses instead the approximation Ê, where Ê 1 N N k=1 | y(k) |2 . As the number of samples N becomes large, by the law of the large numbers, Ê converges to E. The ED is one of the simplest signal detectors. Its operation is very straightfor- ward, and it has a very easy implementation, because it depends only on simple and readily available information. Nevertheless, in spite of its simplicity, the ED is not a perfect solution. The approximation of signal energy E gets better as N increases. Thus, the performance of the ED is directly linked to the number of samples. Furthermore, the ED relies completely on the variance of the noise σ2, which is taken as a fixed value. This is generally not true in practice, where the noise floor varies. Essentially, this means
- 34. Spectrum Sensing in Cognitive Radio Networks ■ 11 ν (a) (b) H0 H1 ν H0 H1 ε̂ ε̂ σ2 Figure 1.2 (a) Ideal ED scheme. (b) Detection uncertainty for the ED. that the ED will generate errors during those variations, especially when the SNR is very low, as shown in Figure 1.2b, where we see an area of uncertainty surrounding the threshold ν in contrast to the case portrayed in Figure 1.2a, in which perfect noise knowledge is considered. 1.3.1.1 Characterization of Energy Detector in AWGN Channels This case has been studied in the work of Urkowitz in 1967 [16]. It is known that the energy detection is the optimal signal detector in AWGN considering no prior information on the signal structure [17]. To understand the inner workings of the ED in this scenario, we need to understand how the probability of detection Pd = Prob{Ê ν|H1} and false alarm Pf = Prob{Ê ν|H0} behave with the measured received signal energy. Take n(k) ∼ NC 0, σ2 as the AWGN noise sample. Then, we know that for the noise-only case, the distribution of the energy of n over T samples can be approximated by a zero mean chi-square distribution χ2 2TW [16], where W is the total bandwidth. Similarly, the energy over T samples of the sum of a signal plus noise can be represented by a noncentral chi-square distribution χ2 2TW (λ) [16], where λ is the noncentrality parameter. Briefly: Ê ∼ χ2 2TW , H0 χ2 2TW (λ), H1 . With these considerations in mind Pf = Qm( λ ξ, √ ν) (1.3) and Pd = (TW , ν/2) (TW ) , (1.4)
- 35. 12 ■ Cognitive Radio Networks where Qm is the Marcum Q-function is the gamma function ξ is the SNR seen by the signal detector 1.3.1.2 Characterization of Energy Detector in Fading Channels In 2002, Kostylev [21] studied the performance of the ED in fading channels. He derived analytical expressions for the ED over the Rayleigh fading channel case (also analyzed the Rice and Nakagami cases numerically). In 2003, the problem was revisited by Digham et al. [22], who provided an alternative analytical development for these three kinds of fading channels. In this section, however, we will restrict the analysis to the more commonly adopted Rayleigh fading. Let us begin by recalling that, in this case, the model of interest is the one shown in Equation 1.14. As such, similar to what Urkowitz did in [16], Kostylev characterized the statistics of the energy of the signal for both the H0 and H1 cases, under the assumption that h(k) is Rayleigh distributed: Ê ∼ χ2 2(TW +1), H0 e2(ξ2+1) + χ2 2TW (λ), H1 , where e2(d2+1) is the exponential distribution with parameter α = 2(ξ2 + 1) with probability density function f (x, α) = αe−αx ξ is the SNR It is clear that, under the hypothesis H0, the statistics are the same as for the AWGN channel case, so the probability of false alarm is the same as in Equation 1.3. Pf = Qm( λ ξ, √ ν). (1.5) The H1 case behaves differently and has the probability of detection given by [22] Pd = e Ê 2 TW −2 m=0 1 m! Ê 2 + 1 + ξ ξ TW −1 e Ê 2(1+ξ) − e Ê 2 TW −2 m=0 1 m! Êξ 2(1 + ξ) . (1.6) 1.3.2 Matched Filter Detector We have seen previously in Section 1.3.1 that the best sensing technique in an AWGN environment, and without any knowledge of the signal structure, is the
- 36. Spectrum Sensing in Cognitive Radio Networks ■ 13 ED. If we do assume some knowledge of the signal structure, then we can achieve a better performance. Most of the wireless technologies in operation include the transmission of some sort of pilot sequence to allow channel estimation, to beacon its presence to other terminals, and to give a synchronization reference for subsequent messages. Sec- ondary systems can exploit pilot signals to detect the presence of transmissions of primary systems in their vicinity. If a pilot signal is known, then the MF signal detector achieves the optimal detection performance in AWGN channel, since it maximizes the SNR, as shown by Tandra and Sahai in [17]. Let us assume that ■ The signal detector knows the pilot sequence x(k), the bandwidth, and the center frequency in which it will be transmitted. ■ The pilot sequence is always appended to the transmission of each primary system (uplink or downlink). ■ The signal detector can always receive coherently. Then, if y(k) is a sequence of received samples at instant k ∈ {1, 2, . . . , N} at the signal detector, the decision rule can be stated as [25] decide for H0, if Ŝ ν H1, if Ŝ ≥ ν , where Ŝ = N k=1 y(k)x(k)∗ (1.7) is the decision criterion ν is the threshold to be compared x(k)∗ is the transpose conjugate of the pilot sequence Here the threshold ν is not the noise variance as it was for the ED. The hypothesis decision is simplified as the MF maximizes the power of Ŝ as shown in Equation 1.7. This means that it performs well even in a low-SNR regime. The MF has some drawbacks. First, a cognitive spectrum sensor might not know which networks are in operation in the environment at a given moment. Therefore, it may not know which sets of pilots to look for. One must remember that if it tries to match an incorrect pilot, it will sense an empty medium and will incorrectly conclude that the medium is free. Second, the MF requires that every medium access be “signed” by a pilot transmission, but this is not the case in general. Furthermore, pilot sequences are only transmitted in the downlink direction. This leaves the
- 37. 14 ■ Cognitive Radio Networks uplink transmissions uncovered. Third, the MF requires coherent reception, which is generally hard to achieve in practice. 1.3.2.1 Characterization of the Matched Filter Signal detection using the MF was studied in 2006 by Cabric et al. in [25]. They showed that Ŝ is Gaussian: Ŝ ∼ N 0, σ2 nε , H0 N ε, σ2 nε , H1 , where σ2 n is the variance of the noise and ε = N k=1 x(k)2 . Based on this information, the probabilities of false alarm Pf and detection Pd are Pf = Q Ŝ εσ2 n (1.8) and Pd = Q Ŝ − ε εσ2 n . (1.9) 1.3.3 Cyclostationary Feature Detection As we have seen, although it performs well, even in the low-SNR regime, the MF requires a good knowledge of the signal structure, which secondary terminals may not have. The natural question to ask is whether we can still be able to perform spectrum sensing with a limited knowledge of the signal structure, perhaps based on a characteristic that is common to most known transmitted signals. In the following text, we show that it is indeed possible. The cyclostationary feature detector relies on the fact that most signals exhibit periodic features, present in pilots, cyclic prefixes, modulations, carriers, and other repetitive characteristics [6,33–37]. Because the noise is not periodic, the signal can be successfully detected. The works by Gardner [33] in 1991 and Enserink et al. [34] in 1995 have studied this signal detection scheme in detail. The work of Enserink et al. follows the same line of the one by Gardner, in which the cyclostationary feature detector is based on
- 38. Spectrum Sensing in Cognitive Radio Networks ■ 15 the squared magnitude of the spectral coherence, which for any random process X is given by |ρα X ( f )| = |Sα X ( f )|2 SX f + α 2 SX f − α 2 1 2 , (1.10) where SX is the spectral correlation density function α is the cyclic frequency f is the spectral frequency In the specific case of the cyclostationary feature detector, substituting ρα X ( f ) by ρ̂α X ( f ) and SX by ŜX , which are the estimated versions of the same quantities, we have the decision metric: M̂ = |ρ̂α X ( f )| = |Sα X ( f )|2 ŜX f + α 2 ŜX f − α 2 , (1.11) which goes into the decision statistic, given by decide for H0, if M̂ ν H1, if M̂ ≥ ν , A recent work focuses on a cyclostationary feature detector for cognitive radio networks [37], called multi-cycles detector. In this work, a cyclostationarity detector scheme is employed on nonfiltered signals, such as OFDM, to detect the cyclic frequency and its harmonics. Finally, it is thought that the cyclostationary feature detector is the most promising signal detection technique as it combines good performance with low requirements on the knowledge of the signal structure [35]. 1.4 Cooperative Sensing Techniques Although for simple AWGN channels most classical approaches perform well, as we have seen, in the case of fading these techniques are not able to provide satisfactory results due to their inherent limitations and to the hidden node problem. To this end, several works [23–31] have looked into the case in which cooperation is employed in sensing the spectrum. Consider the scenario depicted in Figure 1.3, in which primary users (in white) communicate with their dedicated (primary) base station. Secondary receivers {RX1, RX2, RX3, . . . , RXK } cooperatively sense the channel to identify a white space and exploit the medium. The main idea of the cooperative sensing techniques is that each receiver RXi can individually measure the channel and interact on their
- 39. 16 ■ Cognitive Radio Networks RX1 RX2 RX3 RXK Figure 1.3 Cooperative sensing scenario. findings to decide if the medium is available. The main drive behind this idea is that each secondary receiver will have a different perception of the spectrum, as its channel to the receiver will be different from the other secondary receivers, thus decreasing the chances of interfering with hidden nodes. We will concentrate on the scenario depicted in Figure 1.3, although all sensing techniques presented herein can be also applied to alternative scenarios available in the literature, that is, [38]. The cooperative spectrum sensing can be [31] ■ Centralized, in which a central entity gathers all information from all sec- ondary receivers to make a decision about the medium status, which is then transmitted back to the receivers ■ Distributed, in which the receivers share their information to make their own decision In both these situations, the cooperative spectrum sensing is plagued with one problem: how to report or distribute the measures in a resource-constrained network. In fact, if these measurements are the basis for deciding whether a transmission can be made or not, then it does not make any sense to propagate the measurements before the decision is made. To overcome the problem, one could create a dedicated channel for signaling (such as that in [39]) or use an unregulated band (such as ISM). Other works [23–27,30,31] try to restrict the reporting to the minimum possible (often one bit) to ease the process of distributing this information. Finally, [28] considers a hierarchically structured secondary network, in which the secondary spectrum sensors are the secondary base stations, distributed over the sensing area. These base stations would make use of a backbone with enough bandwidth to distribute the measurements among themselves, irrespective of being a single bit or the actual acquired data. Then, during a white space, the terminals are allowed to transmit. Nevertheless, secondary base stations, as opposed to secondary terminals,
- 40. Spectrum Sensing in Cognitive Radio Networks ■ 17 RX1 RX2 RX3 RX4 Figure 1.4 Cooperative sensing scenario. have more processing power and fewer power constraints so that they can perform the spectrum-sensing task better. It should be noted that both of these approaches have their own target applications; neither can be considered the best approach in every case. Another problem of cooperative spectrum sensing is identifying which sec- ondary receivers offer reliable estimations. Let us consider the situation depicted in Figure 1.4, in which one primary terminal is transmitting data in the uplink channel (with low power) toward its primary base station. Several spectrum sensors {RX1, RX2, RX3, RX4} are monitoring the medium detect its state. In this example, {RX2, RX3, RX4} are in range of the transmitter and can correctly sense its ongoing transmission, but RX1 is not.∗ Thus, when the measures of all of the sensors are gathered, how does one select the individual receivers that are performing a reliable measurement? Without knowing the position of the primary transmitter and the channels between secondary receivers and the primary transmitter, this is a com- plicated task. The work by Mishra et al. [30] looks further into the performance impacts of the lack of reliability. Some works [35,40] discuss about a weighting scheme to give different scales to different secondary receivers based on their chan- nel. Other works [23–27] propose a voting scheme to make a trustworthy decision, even with the presence of doubtful measurements. In the remainder of this section, we explore some of the state-of-the-art cooperative sensing techniques. 1.4.1 Voting-Based Sensing We saw in Section 1.3.1 that, in the low-SNR regime, the ED is highly vulnerable to fading and fluctuations in the level of the noise power. What if, instead of employing ∗ This would also apply to the case where RX1 is shadowed or is in a deep multipath fading.
- 41. 18 ■ Cognitive Radio Networks the ED at one location, we could do the same thing in other locations as well? It is expected that among several secondary receivers, even though some will suffer from fading or imprecisions due to the choice of the threshold, some will be able to correctly sense the medium. This is the main idea behind the collaborative spectrum sensing based on voting, studied in a number of works [23–27]. In the voting spectrum sensing, each secondary receiver RXi uses spectrum sensing to form its own decision, as presented in Section 1.3.1. Consider the vector of all responses r such that r = [r1 r2 r3 . . . rK ] , where ri ∈ {1, 0} is the binary response for each sensor i. After all measurements are gathered, the voting procedure takes place [23–25]: decide for H0, if V = 0 H1, if V ≥ 0 , where V = K k=1 rk. Briefly, the voting schemes select H1 if at least one of the secondary receivers decides for H1, which is known as the OR rule. Although this may seem too pessimistic, as it will favor false alarms, according to [23–25], this already gives improvements over the simple energy detection case even for two users. This is reasonable if we remark that with a high number of sensors, higher the probability of reliable spectrum sensing among secondary receivers will be. The probabilities of detection and false alarm for the cooperative approach are Qf = 1 − (1 − Pf )K (1.12) and Qd = 1 − (1 − Pd)K , (1.13) respectively. The work by Sun et al. [26] revisits this scheme to estimate the reliability of each node. In this scheme, only the nodes with reliable sensing are allowed to report their detection. The reliability measure is based on how close the energy of y(k) is to ν, as shown in Figure 1.5. This work defines two new thresholds, ν1 and ν2, that are used to define a “no decision” region. Thus the decision rule can be stated as decide for H0, if 0 ≤ E ≤ ν1 H1, if E ≥ ν2 .
- 42. Spectrum Sensing in Cognitive Radio Networks ■ 19 ν ν1 ν2 H0 H1 E (|y(k)|2) No detection Figure 1.5 Reliability decision scheme. If E falls in (ν1, ν2), then the secondary receiver decides not to report. This way the overall decision, based on the OR rule, concentrates on the reports of M users with a reliable detection out of K total users. The results from this work suggest an increased performance over the conventional case, where no reliability information is used. Another work by Sun et al. [27] proposes a cluster-based spectrum sensing. In this work, a cluster is a grouping of secondary receivers that are spatially close. In each cluster, one receiver, called the cluster head, is elected to do the local decision and the reporting to the central decision entity. There the final decision takes place. 1.4.2 Correlator-Based Sensing Another possibility is to gather all received samples at a central entity that will take the decision instead of leaving the decision of the medium availability to the secondary receivers. With an overall view of the situation, the central entity can decide how to manage the measurements for the decision-taking task better. The schemes presented in this and the following sections all involve such a central entity. Let us, for simplicity sake, suppose that all secondary receivers {RX1, RX2, RX3, . . . , RXK }, shown in Figure 1.3 are within the range of a certain primary transmitter. Then, considering a flat-faded environment, we have yi(k) = ni(k): H0 hi(k)s(k) + ni(k): H1 , where the subscript i means that each value is to be taken for each user i. We can see that for the H0 hypothesis, all yi(k) are independent because they are only composed of AWGN noise. On the other hand, in the H1 hypothesis, all yi(k) are composed of not only the noise but also the signal component s(k) modulated by the channel hi(k). As we know, the signal is common for all users, because it is broadcast by the primary transmitter. We can exploit this fact to detect the presence of transmitted signals by focusing on the correlation between received signals from secondary receivers.
- 43. 20 ■ Cognitive Radio Networks This correlation is calculated via the cyclic convolution, defined as R(ij)(k) = N k=1 yi(a)yj((k − a) mod N) where i and j are the indices of any two secondary receivers. In this scheme, the decision rule is given by decide for H0, if L ν H1, if L ≥ ν , where L is the decision statistic calculated as L = max (i,j)∈B max k (R(ij)), (1.14) where R(ij) is the pairwise cyclic convolution for all permutation of secondary receivers B = {(x, y) ∈ A × A|y ≥ x + 1}, A = {1, 2, . . . , N}. Note that unlike the MF, this scheme does not require coherent reception, as it looks for the highest correlation between any two pairs of sensors. Nevertheless, in the case of coherent reception, we could rewrite Equation 1.14 as L = N k=1 (R(ij)), ∀ (i, j) | i = j, which would effectively maximize the SNR. As far as the authors know, this spectrum sensing scheme has not yet been studied in the literature and thus its performance is not known. It would likely suffer from the same problem as the MF, namely, the challenge of correctly choosing ν. The main limitation of this scheme would be its necessity to report all the measurements, which would require an infrastructure with a very high bandwidth dedicated for the task. 1.4.3 Eigenvalue-Based Sensing Eigenvalue-based sensing is another technique for cooperative sensing, introduced by Cardoso et al. [28] and Zeng et al. [29], based on evaluating the eigenvalues of a matrix formed by the samples collected by multiple sensors in relation to the Marchenko–Pastur law. Herein, we explore the approach as was presented in [28] because the approach in [29] is very similar. To better understand how this spectrum-sensing procedure works, we start with the following assumption:
- 44. Spectrum Sensing in Cognitive Radio Networks ■ 21 ■ The K base stations in the secondary system share information between them. This can be performed by transmission over a wired high-speed backbone. ■ The base stations are analyzing the same portion of the spectrum. Let us consider the following K × N matrix consisting of the samples received by all the K secondary receivers RXi: Y = ⎡ ⎢ ⎢ ⎢ ⎢ ⎢ ⎣ y1(1) y1(2) · · · y1(N) y2(1) y2(2) · · · y2(N) y3(1) y3(2) · · · y3(N) . . . . . . . . . yK (1) yK (2) · · · yK (N) ⎤ ⎥ ⎥ ⎥ ⎥ ⎥ ⎦ . Then, the objective of the eigenvalue-based approach is to perform a test of independence of the signals received at RXi. As said before, in the H1 case, all the received samples are expected to be correlated, whereas in the H0 case, the samples are decorrelated. Hence, in this case, for a fixed K and N → ∞, under the H0 assumption the sample covariance matrix 1 N YYH converges to σ2I. However, in practice, N can be of the same order of magnitude as K and therefore one cannot infer directly 1 N YYH independence of the samples. This can be formalized using tools from random matrix theory [41]. In the case where the entries of Y are independent (irrespective of the specific probability distribution, which corresponds to H0), we can use the following result from asymptotic random matrix theory [41]: THEOREM 1.1 Consider a K × N matrix W whose entries are independent zero-mean complex (or real) random variables with variance σ2 N and fourth moments of order O 1 N2 . As K , N → ∞ with K N → α, the empirical distribution of W WH converges almost surely to a nonrandom limiting distribution with density f (x) = 1 − 1 α + δ(x) + (x − a)+(b − x)+ 2παx where a = σ2 (1 − √ α)2 and b = σ2 (1 + √ α)2 , which is known as the Marchenko–Pastur law. Interestingly, under the H0 hypothesis, the support of the eigenvalues of the sample covariance matrix (in Figure 1.6, denoted by M̌P) is finite. The Marchenko– Pastur law thus serves as a theoretical prediction under the assumption that the
- 45. 22 ■ Cognitive Radio Networks M̌P a b Figure 1.6 The Marchenko–Pastur support (H0 hypothesis). matrix is “all noise.” Deviations from this theoretical limit in the eigenvalue distribution should indicate nonnoisy components. In the case in which a signal is present (H1), Y can be rewritten as Y = ⎡ ⎢ ⎣ h1 σ 0 . . . ... hK 0 σ ⎤ ⎥ ⎦ ⎡ ⎢ ⎢ ⎢ ⎣ s(1) · · · s(N) z1(1) · · · z1(N) . . . . . . zK (1) · · · zK (N) ⎤ ⎥ ⎥ ⎥ ⎦ , where s(k) and zi(k) = σni(k) are, respectively, the independent signal and noise with unit variance at instant k and secondary receiver i. Let us denote by T the matrix T = ⎡ ⎢ ⎣ h1 σ 0 . . . ... hK 0 σ ⎤ ⎥ ⎦ . TTH clearly has one eigenvalue equal to λ1 = |hi|2 + σ2 and all the rest equal to σ2. The behavior of the eigenvalues of 1 N YYH is related to the study of the eigenvalues of large sample covariance matrices of spiked population models [42]. Here, the SNR ξ is defined as ξ = |hi|2 σ2 . The works by Baik et al. [42,43] have shown that when K N 1 and ξ K N (1.15)
- 46. Spectrum Sensing in Cognitive Radio Networks ■ 23 M̌P a b b΄ Figure 1.7 The Marchenko–Pastur support plus a signal component. (which are assumptions that are clearly met when the number of samples N are sufficiently high), the maximum eigenvalue of 1 N YYH converges almost surely to b = |hi|2 + σ2 1 + α ξ , which is greater than the value of b = σ2(1 + √ α)2 seen in the H0 case. Therefore, whenever the distribution of the eigenvalues of the matrix 1 N YYH departs from the Marchenko–Pastur law, as shown in Figure 1.7, the detector decides that the signal is present. Hence, we apply this feature from a spectrum-sensing point of view. Considering λi as the eigenvalues of 1 N YYH and G = [a, b], the cooperative sensing scheme works in two possible ways. 1.4.3.1 Noise Distribution Unknown, Variance Known In this case, the decision criteria used is decide for H0: if λi ∈ G H1: otherwise. (1.16) 1.4.3.2 Both Noise Distribution and Variance Unknown The ratio of the maximum and the minimum eigenvalues in the H0 hypothesis case does not depend on the noise variance and thus serves well as a criteria independent of the noise decide for H0: if λmax λmin ≤ (1+ √ α)2 (1− √ α)2 H1: otherwise. (1.17) It should be noted that, in this case, one still needs to take a sufficiently high number of samples N such that the conditions in Equation 1.15 are met. In other words, the number of samples scales quadratically with the inverse of the SNR.
- 47. 24 ■ Cognitive Radio Networks Note, moreover, that the test under H1 hypothesis also provides a good estimator of the SNR ρ. Indeed, the ratio of largest eigenvalue (b ) and smallest (a) of 1 N YYH is related solely to ρ and α: b a = (ρ + 1) 1 + α ρ (1 − √ α)2 . 1.5 Conclusions and Open Issues In this chapter, the state of the art of spectrum-sensing techniques for cognitive radio networks were covered. We presented not only the classical techniques, inspired by the signal detection approaches developed for radar systems, but also some newly developed ones, carefully tailored for the cognitive radio network sce- nario. Furthermore, we presented their operation, characteristics, advantages, and limitations. In spite of the popularity of spectrum sensing as a study subject for cognitive radio networks, there are still some open issues in this area. Generally, the study has tackled the sensing techniques themselves but little work has considered the systemic point of view, implementation issues, and the complexity of techniques concerning spectrum sensing. Some open issues can be highlighted: ■ Adaptive spectrum sensing. The techniques for spectrum sensing studied so far consider well-behaved scenarios. For some of these techniques, it is quite clear that time-varying environments would greatly compromise their per- formance. Because cognitive radio networks will most likely operate in such environments, it is important that adaptive spectrum-sensing techniques be devised. ■ Cooperation between primary and secondary systems. Is spectrum sensing the best way to find out the medium availability? In some scenarios, maybe not. It is possible that by sharing some information to spectrum brokers, primary systems may benefit from less, or even zero, interference from secondary systems. ■ Cooperative sensing. It is clear that cooperative sensing may be the best option for spectrum sensing in many faded environments. However, there are still some open issues in this field, such as the impact of imperfect information exchange between secondary receivers. ■ Complexity and implementation issues. One of the main limitations of the cognitive radios, and hence of spectrum sensing, is the physical limitation of the hardware and radio frequency (RF) components required. Today, no one knows how to create these cognitive radio transceivers in production scale, with a small package and consuming low power. Another open question is how to find the right bandwidth size for spectrum sensing. Although wideband
- 48. Spectrum Sensing in Cognitive Radio Networks ■ 25 sensing would give a faster and clearer overall picture of the spectrum, it would provide a very rough estimate, because the sensing energy is distributed over a large spectrum. Sweeping the spectrum with narrowband sensing concentrates the sensing energy, but might be too slow in relation to the fast- changing environments. Furthermore, because it is envisioned that sensing will be done by terminals, how do all sensing techniques compare in terms of implementation complexity, energy usage, and processing power? ■ Cognitive pilot channel. The CPC is a specific frequency channel reserved for the diffusion of cognitive radio-related information, such as current frequency band allocation. This interesting new concept could alleviate the requirements of spectrum sensing and provide better performances. It requires further studies to evaluate its gains over the traditional approach. References 1. J. Mitola. Cognitive radio an integrated agent architecture for software defined radio, PhD thesis. Royal Institute of Technology (KTH), Stockholm, Sweden, May 2000. 2. S. Haykin. Cognitive radio: Brain-empowered wireless communications. IEEE Journal on Selected Areas in Communications, 23(2):201–220, 2005. 3. I.F. Akyildiz, W.Y. Lee, M.C. Vuran, and S. Mohanty. Next generation/dynamic spec- trum access/cognitive radio wireless networks: A survey. Computer Networks, 50(13): 2127–2159, 2006. 4. Spectrum Efficiency Working Group. Report of the spectrum efficiency working group. Technical report, FCC, Washington, DC, November 2002. 5. J. Palicot and C. Roland. A new concept for wireless reconfigurable receivers. IEEE Communications Magazine, 41(7):124–132, 2003. 6. A. Fehske, J. Gaeddert, and J.H. Reed. A new approach to signal classification using spectral correlation and neural networks. In Proceedings of the IEEE International Symposium on New Frontiers Dynamic Spectrum Access Networks, Vol. 1, Baltimore, MD, 2005, pp. 144–150. 7. R. Hachemani, J. Palicot, and C. Moy. A new standard recognition sensor for cognitive radio terminal. In USIPCO 2007, Poznan, Poland, September 3–7, 2007. 8. A. Bouzegzi, P. Jallon, and P. Ciblat. A second order statistics based algorithm for blind recognition of OFDM based systems. In Proceedings of Globecom 2008, New Orleans, LA, 2008. 9. A. Sahai and D. Cabric. Spectrum sensing: Fundamental limits and practical chal- lenges. In Tutorial Presented at the 1st IEEE Conference on Dynamic Spectrum Management (DySPAN’05), Baltimore, MD, 2005. 10. Z. Tian and G.B. Giannakis. A wavelet approach to wideband spectrum sensing for cognitive radios. In 1st International Conference on Cognitive Radio Oriented Wireless Networks and Communications 2006, Mykonos Island, Greece, 2006, pp. 1–5. 11. Y. Hur, J. Park, W. Woo, K. Lim, C.H. Lee, H.S. Kim, and J. Laskar. A wideband analog multi-resolution spectrum sensing (MRSS) technique for cognitive radio (CR)
- 49. 26 ■ Cognitive Radio Networks systems. In Proceedings of the IEEE International Symposium on Circuits and Systems 2006 (ISCAS’06 ), Island of Kos, Greece, 2006, p. 4. 12. Z. Quan, S. Cui, A.H. Sayed, and V. Poor. Wideband spectrum sensing in cognitive radio networks. arXiv:0802.4130, 2008. 13. M.I. Skolnik. Introduction to Radar Systems. McGraw-Hill, Singapore, 1980. 14. H.L. Vantrees. Detection, Estimation and Modulation Theory, Vol. 1. Wiley, New York, 1968. 15. V. Poor. An Introduction to Signal Detection and Estimation. Springer, New York, 1994. 16. H. Urkowitz. Energy detection of unknown deterministic signals. Proceedings of the IEEE, 55:523–531, 1967. 17. R. Tandra and A. Sahai. Fundamental limits on detection in low SNR under noise uncertainty. In International Conference on Wireless Networks, Maui, HI, 2005. 18. J.G. Proakis and M. Salehi. Digital Communications. McGraw-Hill, New York, 1995. 19. S.S. Haykin. Communication Systems. Wiley, New York, 2001. 20. C.L. Fullmer and J.J. Garcia-Luna-Aceves. Solutions to hidden terminal problems in wireless networks. In Proceedings of the ACM SIGCOMM’97 Conference on Applications, Technologies, Architectures, and Protocols for Computer Communication, Cannes, France, 1997, pp. 39–49. 21. V.I. Kostylev. Energy detection of a signal with random amplitude. In Proceedings of the IEEE International Conference on Communications (ICC’02), Vol. 3, New York, April–May 2002. 22. F.F. Digham, M.S. Alouini, and M.K. Simon. On the energy detection of unknown signals over fading channels. In Proceedings of the IEEE International Conference on Communications (ICC’03), Vol. 5, Anchorage, AK, May 11–15, 2003. 23. A. Ghasemi and E.S. Sousa. Collaborative spectrum sensing for opportunistic access in fading environments. In First IEEE International Symposium on New Frontiers in Dynamic Spectrum Access Networks 2005 (DySPAN’05), Piscataway, NJ, 2005, pp. 131–136. 24. A. Ghasemi and E.S. Sousa. Spectrum sensing in cognitive radio networks: The cooperation-processing tradeoff. Wireless Communications and Mobile Computing, 7:1049–1060, 2007. 25. D. Cabric, A. Tkachenko, and R.W. Brodersen. Spectrum sensing measurements of pilot, energy, and collaborative detection. In Proceedings of IEEE Military Communications Conference, Washington, DC, October 2006. 26. C. Sun, W. Zhang, and K.B. Letaief. Cooperative spectrum sensing for cognitive radios under bandwidth constraints. In IEEE Wireless Communications and Networking Conference 2007 (WCNC’07 ), Hong Kong, China, 2007, pp. 1–5. 27. C. Sun, W. Zhang, and K.B. Letaief. Cluster-based cooperative spectrum sensing in cognitive radio systems. In IEEE International Conference on Communications 2007 (ICC’07 ), Glasgow, U.K., 2007, pp. 2511–2515. 28. L.S. Cardoso, M. Debbah, P. Bianchi, and J. Najim. Cooperative spectrum sens- ing using random matrix theory. In International Symposium on Wireless Pervasive Computing (ISWPC’08 ), Santorini, Greece, 2008.
- 50. Spectrum Sensing in Cognitive Radio Networks ■ 27 29. Y. Zeng and Y.C. Liang. Eigenvalue base spectrum sensing algorithms for cognitive radio. arXiV:0804.2960, 2008. 30. S.M. Mishra, A. Sahai, and R.W. Brodersen. Cooperative sensing among cognitive radios. In IEEE International Conference on Communications 2006 (ICC’06 ), Vol. 4, San Francisco, CA, 2006, pp. 1658–1663. 31. G. Ganesan and Y. Li. Cooperative spectrum sensing in cognitive radio networks. In First IEEE International Symposium on New Frontiers in Dynamic Spectrum Access Networks 2005 (DySPAN’05), Baltimore, MD, 2005, pp. 137–143. 32. A. Sahai, N. Hoven, and R. Tandra. Some fundamental limits on cognitive radio. In Allerton Conference on Communication, Control, and Computing, Monticello, IL, 2003. 33. W.A. Gardner. Exploitation of spectral redundancy in cyclostationary signals. IEEE Signal Processing Magazine, 8(2):14–36, 1991. 34. S. Enserink and D. Cochran. A cyclostationary feature detector. In Conference Record of the Twenty-Eighth Asilomar Conference on Signals, Systems and Computers 1994, Vol. 2, Pacific Grove, CA, 1995. 35. D. Cabric, S.M. Mishra, and R.W. Brodersen. Implementation issues in spectrum sensing for cognitive radios. In Conference Record of the Thirty-Eighth Asilomar Con- ference on Signals, Systems and Computers 2004, Vol. 1, Pacific Grove, CA, 2004, pp. 772–776. 36. H. Tang. Some physical layer issues of wide-band cognitive radio systems. In Pro- ceedings of IEEE International Symposium on New Frontiers Dynamic Spectrum Access Networks, Vol. 1, Baltimore, MD, 2005, pp. 151–159. 37. M. Ghozzi, M. Dohler, F. Marx, and J. Palicot. Cognitive radio: Methods for the detection of free bands. Comptes Rendus Physique, 7:794–804, September 2006. 38. S.N. Shankar, C. Cordeiro, and K. Challapali. SpectrumAgile radios: Utilization and sensing architectures. In First IEEE International Symposium on New Frontiers in Dynamic Spectrum Access Networks, Baltimore, MD, 2005, pp. 160–169. 39. P. Codier, D. Bourse, D. Grandblaise, K. Moessner, J. Luo, C. Kloeck et al. Cognitive pilot channel. In Proceedings of WWRF15, Vol. 8, Paris, France, 2005. 40. X. Huang, N. Han, G. Zheng, S. Sohn, and J. Kim. Weighted-collaborative spectrum sensing in cognitive radio. In Second International Conference on Communications and Networking in China 2007 (CHINACOM’07 ), Shanghai, China, 2007, pp. 110–114. 41. V.A. Marchenko and L.A. Pastur. Distributions of eigenvalues for some sets of random matrices. Math USSR-Sbornik, 1:457–483, 1967. 42. J. Baik, G.B. Arous, and S. Peche. Phase transition of the largest eigenvalue for nonnull complex sample covariance matrices. Annals of Probability, 33(5):1643–1697, 2005. 43. J. Baik and J.W. Silverstein. Eigenvalues of large sample covariance matrices of spiked population models. Journal of Multivariate Analysis, 97(6):1382–1408, 2006.
- 52. Chapter 2 Capacity Analysis of Cognitive Radio Networks Xuemin Hong, Cheng-Xiang Wang, John Thompson, and Hsiao-Hwa Chen Contents 2.1 Classification of Cognitive Radio Networks .......................... 30 2.1.1 Noninterfering CR Networks ................................. 30 2.1.2 Interference-Tolerant CR Networks .......................... 31 2.1.3 Central Access CR Networks .................................. 32 2.1.4 Ad Hoc CR Networks ......................................... 33 2.1.5 Capacity Analysis: The State of the Art and Motivation ...... 33 2.2 Transmit Power Control .............................................. 35 2.3 Capacity Analysis of a Central Access Cognitive Radio Network .... 39 2.3.1 System Model.................................................. 39 2.3.2 Capacity Analysis and Numerical Results ..................... 39 2.4 Capacity Analysis of a Cooperative Cognitive Radio Network ....... 41 2.4.1 System Model.................................................. 41 2.4.2 Cooperative Communications and Signaling ................. 41 29
- 53. 30 ■ Cognitive Radio Networks 2.4.3 Capacity Analysis .............................................. 44 2.4.4 Results and Discussions........................................ 46 2.5 Conclusions and Open Issues ......................................... 50 2.6 Appendix: Derivation of (2.4) through (2.9) ......................... 51 References ................................................................... 52 Current static and rigid spectrum licensing policy has resulted in very inefficient spec- trum utilization [1–3]. Cognitive radio (CR) [4–8] has been extensively researched in recent years as a promising technology to improve spectrum utilization. The ultimate goal of CR research is to establish a CR network that is either self-sufficient in delivering a multitude of wireless services or capable of assisting existing wire- less networks to enhance their performance. The performance of a CR network is inevitably affected by the coexisting primary systems. Most importantly, the CR transmissions should be carefully controlled to guarantee that the primary services are not jeopardized. To better understand the ultimate performance limits and potential applications of CR networks, it is crucial to study the CR network capacity to provide theoretical insights into the CR network design. In this chapter, we first introduce the classifications of CR networks. We then analyze the capacities of two promising CR networks under average interference power constraints. The first one is a central access CR network, which aims to provide broadband access to CR devices with central base stations (BSs). The second one is a cooperative CR network, where multiple dual-mode CR-cellular users collaborate in the CR band to improve the access performance in the cellular band. Under a simple power control framework, the uplink channel capacities of both CR networks are analyzed and compared, taking into account various system-level factors such as the densities and locations of primary/CR users and path loss in radio propagation channels. Finally, the chapter concludes with some open research issues. 2.1 Classification of Cognitive Radio Networks The core of a CR network is a coexistence mechanism that controls the spectrum sharing in such a way that the operations of the primary system are not com- promised. Based on different coexistence methods, CR networks can be classified into noninterfering CR networks [9–14] and interference-tolerant CR networks [14–17]. On the other hand, based on different radio access types [6], CR networks can be classified as central access/infrastructure-based CR networks [9,12,13] and ad hoc CR networks [18]. In what follows, we briefly explain these four types of CR networks. 2.1.1 Noninterfering CR Networks Noninterfering CR networks exploit the existence of underutilized spectrum, which refers to the frequency segments that have been licensed to a particular primary
- 54. Capacity Analysis of Cognitive Radio Networks ■ 31 service, but are completely unused or partly utilized at a given location or a given time. The unused frequency segments are also called frequency voids, spectrum holes, or white spaces [7], while the partly used spectra are often referred to as grey spaces. A noninterfering CR network seeks to collect these underutilized spectra and reuse them on an opportunistic basis. With careful design, a noninterfering CR network can coexist well with the primary system because it essentially seeks to operate in a signal space orthogonal to the primary signals. A number of measurement campaigns have shown that a large amount of white space exists in two frequency bands: 400–800 MHz and 3–10 GHz. Therefore, noninterfering CR might start to operate first in these two bands in the near future. The concept of noninterfering CR networks has been widely accepted and studied, for example, in [9–14], due to its two obvious advantages. First, the “noninterfering” philosophy means that the primary networks can be well pro- tected. Second, the implementation of a noninterfering CR is relatively simple. Typically, a noninterfering CR is an intelligent wireless device that can dynamically sense the radio spectrum, locate unused or underutilized spectrum segments (or wireless channels) in a target spectrum pool, and automatically adjust its transceiver parameters to communicate in the discovered free channels. Such a sensing-based approach allows minimum changes to the primary system to tolerate CR networks. The IEEE 802.22 working group is currently developing the first wireless standard [9,10] based on the noninterfering CR networks. The aim is to construct a fixed point-to-multipoint wireless regional area network (WRAN) utilizing white spaces in the TV frequency band between 54 and 862 MHz. 2.1.2 Interference-Tolerant CR Networks The interference-tolerant CR networks allow CR users to operate on frequency bands assigned to the primary system as long as the total interference power received at the primary receivers remains below a certain threshold [14–17]. As a new metric to assess the interference at primary receivers, the concept of interference temperature [2] was proposed by the Federal Communications Commission (FCC) in 2002. Similar to the concept of noise temperature, interference temperature measures the power and bandwidth occupied by interference. Moreover, the concept of interference temperature limit [2] was introduced to characterize the “worst-case” interfering scenario in a particular frequency band and at a particular geographic location. CR transmissions in a given band are considered to be “harmful” only if they would raise the interference temperature above this limit. Unlike traditional transmitter-centric approaches that seek to regulate interference indirectly by controlling the emissions of interfering transmitters, the interference temperature concept takes a receiver- centric approach and aims to directly manage interference at primary receivers. Recently, in 2007, the FCC has abandoned its use of “interference temperature” due to current difficulties in implementing this concept. However, the philosophy
- 55. Other documents randomly have different content
- 56. for stoppages, taking up and putting down passengers, which lost many minutes in a journey, and the heavy loads carried, by neither of which was the Old Times troubled, I think the Brighton feat, good as it was, has often been surpassed. The three Birmingham Tally-ho's generally had a spurt on the first of May, and more than once performed the journey of a hundred and eight miles under seven hours—the best record, I believe, in existence. Pace, however, at last, is a relative thing, and eight or nine miles an hour on one road may be really as fast as twelve or thirteen on another. I can safely say that, though I have driven some fast coaches in my time, I never had a day of harder work to keep time than in doing eighty miles in ten hours. What with one weak team in the early part of the journey, hilly roads, a heavy load, and frequent delays for changing passengers and luggage, the last stage of nine miles had to be covered in forty-two minutes to bring us in to time and catch the train. Before finally bidding adieu to the subject of driving, it may perhaps be allowed me to say a few words about harness and the fitting of it. Of course it hardly needs saying that a coachman ought to be familiar with every strap and buckle of it, though this intimate knowledge may be dispensed with by those who only drive their own teams, and are always waited on by one or two good and experienced servants. Indeed, from what I witnessed in Hyde Park several years ago, I have had my suspicions whether these same servants are not sometimes utilised on early mornings in training the teams, and putting them straight for the masters' driving in the afternoon. I once saw a drag brought round to the right at the Magazine without the gentleman in charge of the box touching the off-side reins with his right hand at all; and I fail to see how this could have been accomplished unless the horses were as well trained to it as circus steeds.
- 57. Still, however perfect these men may be as gentlemen's servants, their experience has not generally led them to attend very closely to the exact fitting of the harness—the collars particularly—which used often to be the plague of their lives to stage coachmen, and even might give trouble to a gentleman, if driving an extended tour. A few hints, therefore, from an old hand may perhaps not be thrown away. With horses freshly put into harness their shoulders are always liable to be rubbed, and they require the greatest care and attention; and one thing should always be insisted on in these cases, which is to wash the shoulders with cold water after work, and to leave the collars on till they have become quite dry again. But if care is necessary in the case of gentlemen's work, what must have been that required with coach horses—especially if running over long stages, with heavy loads and in hot weather. Of course, a good deal depended upon the care of the horse-keeper; but nothing he could do had any chance of keeping the shoulders sound if the collars wobbled which they certainly always will do if the least light can be seen between the collar and the upper part of the horse's neck. Then, again, it is most important for the collar to be the right length to suit the individual horse. One which carries his head high will require a longer one in proportion than one which carries it low, because the former position of the head has the effect of causing the windpipe to protrude. On stage-coach work we never cared so much about the weight of the collar as the fitting, and offering a fairly broad surface to the pressure. Two or three pounds extra weight in a collar is nothing compared to the comfortable fitting of it, as we ourselves know to be the case with half-a-pound or so when walking a long distance in strong boots. If a wound should appear, after all the care that can be taken, a paste made of fullers' earth with some weak salt and water will nearly always effect a cure, if the collar is properly chambered, so as to remove all pressure from the part. In case of a shoulder showing a disposition to gall, I always carried in the hind boot two or three
- 58. small pads, which I could strap on to the collar, so as to remove the pressure temporarily till it could be chambered; and any gentleman embarking on a driving tour would find this to be a good precaution to take, especially if he is going into out-of-the-way districts. I will conclude in the words of Horace— Si quid noviste rectius istis, Candidus imperti: si non his utere mecum.
- 59. CHAPTER XXIII. THE END OF THE JOURNEY. And now, ladies and gentlemen, I leave you here, and trust I have given you no cause for complaint on the score of either civility or politeness to my passengers. I fear that in some places the road may have been heavy and the pace slow. Perhaps it may be thought that the style is incoherent, to which I can only say that such is usually the character of chatter; and if I have written anything which has afforded some interest or amusement, my most ardent hopes are satisfied. The tale I have told has, in one sense, been told before, but so many fresh phases and incidents were so constantly turning up in the old mode of travelling, that it is not necessarily a twice-told tale. Probably the first idea of most readers upon closing the book will be, How thankful I am that my lot was not cast in the days of my father or grandfather; and this naturally leads to the reflection that when the busy wit of man had not produced so many inventions for evading the minor ills of life, the first idea was to endure them; but now, when fresh schemes of all sorts and descriptions are being propounded every day to render life easy, it is to cure them; and if this does not go to the length of making artificial wants, no doubt it is the wisest course to adopt. To the old hand, however, who has not forgotten his early experiences, this eagerness to escape all hardship may seem to savour of softness and effeminacy, but I make no doubt that, though not called forth as it used to be in the days of yore, there still exists in the youth and manhood of Old England the same pluck and power of endurance when duty calls, as there ever was; and that as long as
- 60. we continue to cherish our old field sports and games, we are not in much danger of losing them. It were folly to stand up for road travelling as against the greater convenience of railways; still, I confess to a lingering feeling of regret that what was brought to such a state of perfection should have so completely vanished, and I think I cannot express these feelings better than by a short anecdote. Many years ago, when hunting with the late Sir W. W. Wynn's hounds, when they had the advantage of the guidance of John Walker, I asked him which pack, whether the large or small, showed the best sport and killed the most foxes. His answer was, Well, I really think the large pack does kill most foxes and give the best sport altogether, but I like the little ones. And if asked which is the best mode of travelling, whether by road or rail, I must confess that, as a travelling machine for conveying us from one part of the country to another, the railway is the best both for safety, speed, and economy; but having said this, I am constrained to make the same sort of reservation as was made by John Walker, and say, I like the coaches. Most noticeable of all, perhaps, was the plucky effort made in 1837 to revive the favourite Red Rover coach between London and Manchester, which had been discontinued upon the opening of the London and Birmingham and the Grand Junction Railways. It was the last charge of the Old Guard, and shared the same fate. It may be interesting, however, to append a copy of this singular notice— one more evidence of the reluctance of Englishmen to be beaten, even at long odds. The very date at foot is significant, for the enterprise was embarked on in the teeth of the approaching winter.
- 61. An old song may come in here:—
- 62. The road, the road, the turnpike road, The hard, the brown, the smooth, the broad, Without a mark, without a bend, Horses 'gainst horses on it contend. Men laugh at the gates, they bilk the tolls, Or stop and pay like honest souls. I'm on the road, I'm on the road, I'm never so blithe as when abroad With the hills above and the vales below, And merry wheresoe'er I go. If the Opposition appear in sight, What matter, what matter, we'll set that all right. In the introduction I ventured to point out some inaccuracies which I had observed in a statement made upon the subject of coach fares, and as it is probably one which few remember anything about, I give a statement of what would be about the profit and loss of a month's working of a coach for a hundred miles. RECEIPTS. A Full Load on the Way-bill both ways. £ s. d. 8 inside passengers 15 0 0 14 outside 25 4 0 Parcels 1 0 0 Parcels £ 41 4 0 Month's receipts 988 16 0 Deduct expenses 113 14 0 £ 875 2 0 PAYMENTS.
- 63. Daily £ s. d. 15 toll-gates, at 3s.[3] 2 5 0 Hire of coach, per mile 2½d. 1 0 10 Mileage duty, 2d.[4] 0 6 8 Washing and oiling coaches 0 2 0 £ 4 8 6 For 4 weeks £ 106 4 0 Monthly. 8 road booking-offices £ 4 0 0 2 end booking-offices 2 0 0 Making Share bills 1 0 0 Oil and trimming lamps, say 0 10 0 Total £ 113 14 0 This makes £8 15s. to be divided per mile, which, of course, would give a very handsome profit; but full loading could not be expected every day, and if it was reduced to half loads, it would not be such a very fat concern. The cost of each horse was usually put at 17s. 6d. a week, including blacksmith, and that, supposing a man to cover a ten-mile stage for which eight horses would be ample if not running on Sundays, would cost £7 a week, or £28 a month, leaving, at about half loading, say £20 profit. But from this has to be deducted saddler, veterinary surgeon, and wear and tear, the two latter of which depend, to a certain extent, on circumstances over which he has not much control, as it depends upon such things as sickness in the stables and accidents. [3]
- 64. It was usual for coaches to come to terms with the pikers to pay for three horses instead of four. [4] There had also to be paid £5 licence duty yearly when the plates were taken out.
- 65. [APPENDIX.] G. P. O. APPENDIX. LIST OF MAIL COACHES WHICH WORKED OUT OF LONDON. Bath, through { Hounslow, } From the { Maidenhead, } Spread Eagle, { Reading, } Gracechurch Street, { Newbury, } and { Hungerford, } Swan with Two { Marlborough, } Necks, { Devizes, } Lad Lane. Birmingham, through { Aylesbury, } From the { Bicester, } King's Arms,
- 66. { Banbury, } Holborn Bridge. { Leamington, } { Warwick, } Brighton, through { Croydon, } From the { Reigate, } Blossoms Inn, { Crawley, } Lawrence Lane. { Cuckfield, } Bristol, through { Hounslow, } From the { Reading, } Swan with Two { Newbury, } Necks, { Marlborough, } Lad Lane. { Calne, } { Chippenham, } { Bath, } Carlisle—See Glasgow. Chester, through { Barnet, } From the { St. Albans, } Golden Cross, { Dunstable, } Charing Cross. { Northampton, } { Hinckley, } { Atherstone, } { Lichfield, } { Stafford, } { Nantwich, } { Tarporley, }
- 67. Devonport, through { Hounslow, } From the { Bagshot, } Swan with Two { Basingstoke, } Necks, { Andover, } Lad Lane. { Salisbury; } { Sherborne; } { Chard, } { Honiton, } { Exeter } Dover, through { Dartford, } From the { Rochester, } Swan with Two { Sittingbourne, } Necks, { Faversham, } Lad Lane. { Canterbury, } Edinburgh, through { Ware, } From the { Buntingford, } Bull and Mouth, { Royston, } St. Martin's-le- Grand. { Caxton, } { Huntingdon, } { Grantham } { Newark } { Doncaster } { Ferry Bridge, } { York, } { Northallerton, } { Darlington, } { Durham, } { Newcastle, }
- 68. { Alnwick, } { Berwick, } { Dunbar, } { Haddington, } Exeter, through { Basingstoke, } From the { Andover, } Bull and Mouth, { Salisbury, } St Martin's-le- Grand. { Blandford, } { Dorchester, } { Bridport, } { Axminster, } { Honiton, } Glasgow, through { Barnet, } From the { Hatfield, } Bull and Mouth, { Baldock, } St Martin's-le- Grand. { Biggleswade, } { Stilton, } { Stamford } { Grantham, } { Newark, } { Doncaster, } { Wetherby, } { Boroughbridge, } { Greta Bridge, } { Appleby, } { Carlisle, }
- 69. Gloucester, through { Hounslow, } From the { Maidenhead, } Cross Keys, { Henley, } Wood Street, { Nettlebed, } and { Oxford } Golden Cross, { Witney, } Charing Cross. { Burford, } { Cheltenham, } Halifax, through { Barnet, } From the { Woburn, } Swan with Two { Newport-Pagnel, } Necks, { Market Harborough, } Lad Lane, { Nottingham, } and { Sheffield, } Bull and Mouth, { Huddersfield, } St. Martin's-le- Grand. Hastings, through { Farnborough, { From the { Tunbridge, { Golden Cross, { Lamberhurst, { Charing Cross. { } and Bolt in Tun, { } Fleet Street. Holyhead, through { Barnet, } From the { St. Albans, } Swan with Two { Coventry, } Necks, { Birmingham, } Lad Lane. { Wolverhampton, } { Shrewsbury, }
- 70. { Oswestry, } { North Wales, } Hull, through { Barnet, } From the { Hertford, } Spread Eagle, { Biggleswade, } Gracechurch Street, { Stilton, } and { Peterborough, } Swan with Two { Folkingham, } Necks, { Lincoln, } Lad Lane. { Brigg, } { Across the Humber to } { Kingston-upon- Hull } Leeds, through { Barnet, } From the { Bedford, } Bull and Mouth, { Higham Ferrers, } St. Martin's-le- Grand. { Kettering, } { Nottingham, } { Sheffield, } { Wakefield, } Liverpool, through { Barnet, } From the { St. Albans, } Swan with Two { Coventry, } Necks, { Lichfield, } Lad Lane. { Newcastle-u-Lyne, }
- 71. { Knutsford, } { Warrington, } Louth, by Boston, through { Caxton, } From the { Peterborough, } Bell and Crown, { Deeping, } Holborn, and { Spalding, } Saracen's Head, { Spilsby, } Skinner Street. Manchester, through { Barnet, } From the { St. Albans, } Swan with Two { Dunstable, } Necks, { Northampton, } Lad Lane. { Market Harborough, } { Leicester, } { Derby, } { Ashbourne, } { Congleton, } { Macclesfield, } Norwich, by Ipswich, through { Ilford, } From the { Romford, } Spread Eagle, { Brentwood, } Gracechurch Street. { Chelmsford, } { Witham } { Colchester, }
- 72. Norwich, by Newmarket, through { Epping, } From the { Bury St. Edmunds, } Belle Sauvage, { Thetford, } Ludgate Hill. Portsmouth, through { Kingston, } From the { Esher, } White Horse, { Guildford, } Fetter Lane and { Godalming, } Bolt in Tun, { Petersfield, } Fleet Street. Southampton and Poole, through { Hounslow, } From the { Staines, } Swan with Two { Bagshot } Necks, { Alton, } Lad Lane, and { Alresford } Bell and Crown, { Winchester, } Holborn. Stroud, through { Hounslow, } From the { Henley, } Cross Keys, Wood { Abingdon, } Street, and the { Faringdon, } Swan with Two Necks, { Cirencester, } Lad Lane. Wells (Norfolk), through { Lynn, } From the { Ely, } Swan with Two { Cambridge, } Necks, { Royston, } Lad Lane.
- 73. { Ware, } Worcester, through { Uxbridge, } From the { Beaconsfield, } Bull and Mouth, { High Wycombe, } St. Martin's-le- Grand. { Oxford, } { Woodstock, } { Chipping Norton, } { Moreton-in-Marsh, } { Evesham, } { Pershore, } Yarmouth, through { Romford, } From the { Chelmsford, } White Horse, { Witham, } Fetter Lane. { Colchester, } { Ipswich, } { Saxmundham, } { Lowestoft, } So much for the main arteries, but the account would hardly be complete without showing how the more remote and out-of-the-way districts were provided for. I will, therefore, add the routes of a few mails which might be considered as prolongations of some of those already mentioned, but they were worked under fresh contracts and with fresh coaches. South Wales was served by three—one from Bristol and two from Gloucester, as shown below:— Bristol to Milford Haven, by { New Passage Ferry, { Newport,
- 74. { Cardiff, { Cowbridge, { Neath, { Caermarthen. Gloucester to Milford Haven, by { Ross, { Monmouth, { Abergavenny, { Brecon, { Llandovery, { Caermarthen, { Haverfordwest. Gloucester to Aberystwith, by Ross, Hereford, Kington, Rhayader, and Dyffryn Castle. The Gloucester and Milford was, I think, driven out of Gloucester at one time by Jack Andrews, a very good coachman, and over the lower ground there was a man of the name of Jones. I may, perhaps, be told that that is not a very distinguishing mark of a man in those parts, perhaps it is not, but if the name failed to convey a knowledge of who he was, he, at any rate, possessed one very characteristic feature which was that he always drove without gloves whatever might be the state of the weather. If he saw his box passenger beating his hands against his body or going through any other process with the vain hope of restoring the circulation into his well-nigh frozen fingers, his delight was to hold out his gloveless hand and say, Indeed, now there is a hand that never wore a glove. And this recalls to my memory another anecdote which was told me a great many years ago, and which, though it refers to the other extremities, may not be inappropriately introduced here. It appertains to a very well known character already mentioned, the
- 75. well known Billy Williams, often spoken of as Chester Billy. I am aware that tales are sometimes engrafted on remarkable characters which are also told of others, still I believe I shall not be doing a wrong to any one if I tell this as 'twas told to me, of our old friend Billy. At any rate, it is too good to be lost, so here it is. On one very cold winter morning it happened that Billy had a box passenger who was stamping his feet on the footboard in the vain attempt to restore the circulation of the blood, which led Billy to remark, Your feet seem cold this morning, sir, to which the gentleman answered, I should think they were, are not yours? No, says Billy, they're not; adding, I expect you wash 'em. Wash them, says the passenger, of course I do, don't you? No, was the reply, I should think not, I iles 'em. The Manchester mail was also prolonged to Carlisle, though the direct Carlisle mail went by a rather shorter route, but then the populous district on the west coast had to be provided for. It travelled through Preston, Lancaster, Kendal and Penrith. This was, over some of the ground at any rate, one of the fastest mails in England. Again, in addition to these, which may be said to have had their origin in London, there existed a considerable number of what were called cross country mails, some of which ran long distances and at high speed, connecting together many important districts. A few of them I will mention, beginning with the Bristol and Liverpool, which was a very fast one. Bristol to Liverpool, by { Aust Passage Ferry, { Monmouth, { Hereford, { Shrewsbury, { Chester, { Woodside Ferry.
- 76. Bristol to Oxford, by { Bath, { Tetbury, { Cirencester, { Fairford, { Faringdon. Liverpool to Hull, by { Warrington, { Manchester, { Rochdale, { Halifax, { Bradford, { Leeds, { Tadcaster, { York. Bristol to Birmingham, by { Gloucester, { Wincanton, { Droitwich, { Bromsgrove. Birmingham to Sheffield, by { Lichfield, { Derby, { Chesterfield. And no doubt there were several others in one part of the country or another, but I have been unable to meet with any regular list of them, though it is very unlikely that such a road as that between Bristol and Exeter by Taunton, for example, should have been left out. This road certainly had a fast coach on it. The Royal Exeter ran from Cheltenham to Exeter through Gloucester and Bristol, driven between Cheltenham and Bristol at one time by Capt. Probyn,
- 77. and afterwards by William Small. It was a fast coach, stopping for dinner at Nisblete's, at Bristol, and then proceeding on its journey to Exeter. Then, again, there was a populous and important district through the Staffordshire Potteries, from Birmingham to Liverpool and Manchester, which must have been provided for somehow, but it is not impossible that this may have been effected by the bags being conveyed to Lichfield by the Sheffield, and then transferred to the down Liverpool and Chester mails. There were also running short distances what were called third class mails, which carried twelve passengers, and the coachman was in charge of the bags. On one of them which ran between Shrewsbury and Newtown I did a good deal of my early practice. And now, having given a list, more or less perfect, of the mails which traversed England and Wales, perhaps a few words on the subject of the pace at which they travelled may not be without interest. After singling out the London and Birmingham day mail, which was timed at twelve miles an hour, it is impossible to say, at the present date, which was the fastest coach. That the Quicksilver was the fastest mail, I have no doubt, though I believe the palm has been disputed by the Bristol, and perhaps some others; for if a passenger asked a coachman which was the fastest, he was very likely to be told that the one he was travelling in was. I cannot, however, believe that any of these claims could have been supported by facts. Cui bono? We can see at a glance why the Devonport should be pushed along as fast as possible, because the journey was a long one; but the distance to Bristol was only one hundred and twenty miles, and whether the mail arrived there at eight or nine o'clock in the morning would have been thought little of in those days, but in a journey of two hundred and twenty-seven miles half a mile an hour makes an appreciable difference. It would seem reasonable, therefore, that the longer mails should have been accelerated as
- 78. much as possible, and so I believe it really was the case, and that the Holyhead was, after the Quicksilver, the fastest out of London. At any rate, I know that, when travelling by it, we always passed all the other mails going the same road, and that included a considerable number, as the north road and the Holyhead were synonymous as far as Barnet, and, moreover, the Post-Office was likely to have screwed up these two mails the tightest, as one carried the Irish bags and the other had the correspondence of an important dockyard and naval station. To single out the fastest coach would be still more impossible. The Wonder had a world-wide reputation, which was well deserved, both for the pace and regularity with which she travelled and the admirable manner in which she was appointed in every way; but what gave that coach its preponderating name was the fact of its being the first which undertook to be a day coach over a distance much exceeding one hundred and twenty miles. The Manchester Telegraph must have surpassed the Wonder in pace, and, certainly, when we consider the difference of the roads and the hills by which she was opposed in her journey through Derbyshire, had the most difficult task to accomplish; and, again, the Hirondelle was timed to go the journey of one hundred and thirty-three miles between Cheltenham and Liverpool in twelve hours and a half, which is a higher rate of speed than the Wonder, which was allowed fifteen and a half hours to cover the one hundred and fifty-four miles between London and Shrewsbury, and on a far superior road. I have been induced to enter into this subject because one sometimes now-a-days meets with people who appear to have a somewhat hazy idea about it, and talk glibly of twelve miles an hour as if it was nothing so very great after all. Well, I am not going to deny that it can be done, because I know that it has been effected by the Birmingham day mail, as already stated, and I have also been told by an old inspector of mails that in the latter days they did contrive to screw some Scotch mails up to that speed; but I am sure
- 79. Welcome to our website – the perfect destination for book lovers and knowledge seekers. We believe that every book holds a new world, offering opportunities for learning, discovery, and personal growth. That’s why we are dedicated to bringing you a diverse collection of books, ranging from classic literature and specialized publications to self-development guides and children's books. More than just a book-buying platform, we strive to be a bridge connecting you with timeless cultural and intellectual values. With an elegant, user-friendly interface and a smart search system, you can quickly find the books that best suit your interests. Additionally, our special promotions and home delivery services help you save time and fully enjoy the joy of reading. Join us on a journey of knowledge exploration, passion nurturing, and personal growth every day! ebookbell.com