Sensing Time Improvement Using Two Stage Detectors for Cognitive Radio System

International Journal of Computer Networks & Communications (IJCNC) Vol.16, No.1, January 2024
DOI: 10.5121/ijcnc.2024.16105 71
SENSING TIME IMPROVEMENT USING TWO STAGE
DETECTORS FOR COGNITIVE RADIO SYSTEM
Mohamed Khalaf 1*
, Ahmed Fawzi 2†
and Ahmed Yahya3†
1*
Department of Electronics and Communications Engineering, Modern Academy for
Engineering and Technology, Maadi, Egypt
2
Department of Electronics and Communications Engineering, Modern Academy for
Engineering and Technology, Maadi, Egypt
3
Electrical Engineering Department, Faculty of Engineering (Cairo), Al-Azhar
University, Cairo, Egypt
ABSTRACT
Cognitive radio (CR) is a promising technology for both present and future telecommunications to satisfy
the demand of the next generation due to inefficient use of the allocated spectrum. Due to its ability to
utilize the available bandwidth of other wireless communication networks and so enhance its occupancy.
Spectrum sensing (SS) is the key characteristic of the CR system that helps it identify the empty spectrum.
SS has gained a lot of interest recently and it is an active research area since it offers additional
opportunities to secondary users. A broad variety of analytical methods to identify the Primary User's
(PUs) presence have emerged as a result of SS techniques for CRs. Although each approach has its own
benefits, the drawbacks attached to them make an individual implementation of the technique impractical
for usage. To mitigate the drawbacks and maximize the benefits offered by the individual methods, a two
stage detector can be employed for SS. However, the stages method lengthens the time needed to sense the
spectrum and produce a definitive result. In this paper, we propose a two-stage sensing approach, where
the first stage is an Interval Dependent Denoising detector (IDD) and followed by a second stage is Energy
Detection (ED) which provides a large decrease in the mean sensing time compared to different related
two-stage SS approaches The ED method is simple and has a short sensing time, but it performs poorly
when the Signal to Noise Ratio (SNR) is low, Hence it is thought that separating PU activity from noise is
essential for accurate SS. Therefore, we use IDD as a noise reduction technique in the first stage before the
ED stage. The simulation results have been utilized to demonstrate that this technique leads to a large
savings in sensing time as compared to an existing two-stage detection approaches.
Keywords:
Cognitive radio (CR), Spectrum sensing (SS), two stage detection, energy detection (ED), mean detection
time, interval dependent denoising detector (IDD).
1. INTRODUCTION
As wireless communication grows rapidly, it becomes increasingly crowded due to inefficient
bandwidth use across different spectrum bands, which is a one of the major drawbacks of
traditional wireless networks with limited bandwidth deployment constraints. In accordance
with estimation of expected communications-related needs, in the next years, it's predicted that
the spectrum will expand even further. However, the current spectrum shortfall is largely as a
result of ineffective fixed spectrum allocations than physical shortage of spectrum. The
conventional fixed spectrum assignment policy is a big hurdle in the innovation of new
technologies is supported by recent studies of the Federal Communication Commission (FCC)
allowed the unlicensed fixed and personal/portable devices in rural and urban area [1, 2].

72
CR has been recently seemed as an effective method to increase spectrum utilization through
opportunistic spectrum sharing in order to resolve the conflicts between spectrum under-
utilization and congestion. CRs are intelligent communication technology that can recognize their
wireless surroundings and decide when and how to use the available spectrum (white space),
which is not being used by the authorized user. To avoid causing damaging interference to main
users, CRs must be capable of scanning wide frequency ranges and detecting unoccupied bands
[3]. The CR must sense the spectrum's availability before accessing the channel in order to
implement without interfering with the primary signal [4].
The primary purpose of CR is SS to identify PU / licensed user; if PU is not present, spectrum is
then available to Secondary User/cognitive radio user (SU), and this area is known as a "spectrum
hole" or "white space". SS is a process for detecting PU by sensing the surrounding radio
environment [5-7]. The fundamental concerns of SS revolve on two issues: first, SU
communication should not interfere with the primary system, and second, spectrum holes should
be efficiently identified for the desired throughput and Quality of Service (QoS) [8].
There are several techniques for SS that are determined, such as matching filter, ED [3] and
Cyclostationary Feature Detection (CFD) [9, 10]. The received SNR in communication systems
is maximized by the matched filter, so it is the optimal technique. However, a key disadvantage
of a matched filter is that it required previous knowledge of the signal of the PU, including the
type and order of the modulation, the pulse shaping, and the packet format. CD offers reliable SS,
even at low SNRs and it is based on the autocorrelation function, but it is computationally
complicated and necessitates longer sensing time. On the other hand, one-order CD [11, 12] is
conducted in the time domain. The mean characteristic of the PU signal is used in one-order CD
to increase the effectiveness of the channel sensing. Therefore, less computational complexity
and real time operation can be accomplished by the one-order CD. on the other hand, ED is a
simple approach with a short sensing time, but it performs poorly in low SNR conditions. There
are also other additional approaches to detect the spectrum such as maximum-minimum
eigenvalue detectors, which may simultaneously achieve high detecting probability (PD) and low
false-alarm probability (PF) without requiring PU prior information [3].
The detection rate and PF serve as indicators of the sensitivity and selectivity, which together
forms the metrics for spectrum sensing performance. The probability that SU announces that PU
is present when the spectrum is vacant is known as the PF ; in contrast, the probability that SU
announces that PU is present when the spectrum is occupied by PU is known as the PD. The
probability that SU will announce that PU is not present when the spectrum is in use is indicated
by the probability of missed detection. Due to the fact that false alarms decrease spectrum
efficiency and miss detection causes interferes with PU, achieving the highest PD while
maintaining the lowest PF is essential for the best detection performance [13].
According to the IEEE 802.22 standard [14], PUs must be identified within two seconds, and PF
and missed detection must be less than or equal to 0.1 [15]. To guarantee that the spectrum
resources are completely utilized without interfering with PUs' communication, the sensing time
must not exceed 2 seconds [16]. It is not possible to transmit data while also sensing the licensed
spectrum. Therefore, for effective use of spectrum holes, SU must periodically sense the band at
sensing period every Tp seconds. Due to SU's not being aware of its activities through the
sensing interval, or until the next sensing moment, PU transmission may be prevented. As a
result, the sensing period has a major impact on PU performance. Maximizing the sensing period
may enhance SU throughput, but may make PU prevented because PU is rarely sensed. From the
viewpoint of a CR network, SU wants to increase the sensing period and reduce the sensing time
to increase transmission time [17]. To cohabit with the licensed band, the SU must correctly
schedule the sensing period. The SU can increase throughput and minimize PU interference while

73
maintaining sensing reliability by minimizing the sensing time. In CR networks, maximizing PU
detection reliability and reducing sensing time are two key considerations for efficient SS. In this
paper, a two-stage sensing model for spectrum detection is developed to increase sensing
efficiency by increasing PD while staying within the parameters of the PF and obtaining a low
mean sensing time. The rest of the paper is arranged as follows. Section 2 displays a brief
discussion of various related work on two-stage SS methods. The system model of the sensing
approach is shown in Section 3; Section 4 includes simulation results for evaluating the proposed
model performance in terms of detection performance and mean detection time and finally
conclusions are noted in section 5.
2. RELATED WORK
According to the benefits and drawbacks of the methods mentioned above, it can be concluded
that no single technique is perfect enough to be used on a CR device. As a result, it would be
preferable to combine two or more techniques in order to mitigate the drawbacks of the
techniques while utilizing the individual benefits of the techniques. This kind of synergy may be
used to establish algorithms that have quick sensing time even in low SNR situations and can
achieve the required probabilities of detection and false alarms. Thus, recent paper on SS is
concentrating on two-stage spectrum detection. The authors of Ref. [18] suggest a two stage
sensing approach for dynamic spectrum access, which has been studied. The authors divided the
sensing techniques into two stages: coarse and fine sensing. The first stage is coarse sensing
based on ED, and if needed the second stage is fine sensing based on CD. Only if a channel is
identified to be vacant in the coarse stage the fine stage is employed and will take the final
decision .If not, coarse sensing will take the final decision. Another two-stage sensing mechanism
was presented by the authors in Ref. [19]. Multiple Energy Detectors (MED) are performed in the
first stage, where each energy detector (ED) consisting of a single antenna and with a fixed
threshold (MED FT) for making a local binary decision, and if needed, the second stage, which is
constructed of ED with Adaptive Double Threshold (ED ADT), is activated with this technique,
computing complexity is reduced, but sensing time is increased. The authors of Ref. [20] use a
two adaptive local SS technique in which CR can utilize either an ED or a one-order
cyclostationary for SS based on expected SNR, which is determined in advance for available
channels. The proposed technique uses the ED to provide rapid detection at high SNRs. at low
SNRs since ED is unreliable, the proposed technique offers improved detection at the cost of
longer mean detection time.
Two-stage SS approaches are suggested by the authors in Ref. [21]. They used two-stage
detectors in the general case, An ED is the first stage followed by a Covariance Absolute Value
(CAV) in the second stage. The authors used the two stages for all the received signals with
variable SNRs. In a modified method, only when the SNR is higher than a specific threshold the
second stage of the CAV detector is activated. Based on Estimated SNR with Adaptive Threshold
(ESNR ADT), the authors of [22] provided a unique collaborative SS method in CRN. The
proposed scheme evaluates SNR and chooses an ED with either a fixed threshold or an adaptive
threshold to improve detection performance at a fixed PF = 0.1. Based on the estimated SNR
value and the first threshold, the detector is selected. If the estimated SNR value exceeds or is
equal to this threshold, an ED with the second threshold will be employed; else, an ED with ADT
will perform sensing processes. Although this technique decreases mean sensing time less than
CD, it does not take into account the issue of SS failure. In [23], the authors suggested a two-
stage sensing approach for CRs, where the first stage involved ED and the second, CD. The
detection parameters are used in both stages to enhance the PD within the boundaries of the PF,
although it is computationally more complex and needs a longer sensing time. The authors of
Ref. [24] presented an improved two-stage detection (improved-TSD) method for SS. The
improved-TSD technique consists of two stages: the first stage contains two detectors, namely an

74
ED using a single adaptive threshold (ED SAT) and an ED using two adaptive thresholds (ED
TAT), which are arranged in parallel, and the second stage includes a decision device (DD) that
creates the final decision using the OR-rule. The proposed sensing technique reduces the sensing
time.
With the aid of the two stages ED and CD discussed in [18], however the elapsed time by a two
stage detector is still high because of the complexity of the second stage. Better performance
features are presented by the authors In Ref. [25]. As a result, the second stage detector can be
used intelligently and decreasing the time taken to reach a conclusive decision. The decrease in
the mean detection time of the SS algorithm in this method illustrates the advantage of reduction
in the activation of the second stage. It was further shown that the suggested modification has no
bad effect on PD of the CR and the authors did not consider uncertainty in their estimation of the
noise power for the ED stage. In Ref. [26], a hybrid sensing model for spectrum detection in CR
is executed. The first path is constructed from two sequential sensing stages; in the first stage, an
ED is used to identify the presence of the PU signal when the signal has not been detected. The
presence of the PU signal is detected using a second stage called Maximum-Minimum Eigen
value (MME). The second path composed of two parallel stage detectors employing separate ED
and MME to detect the PU signal separately. The two results are brought together to form a
decision, and the final decision is taken using the combined results of the two paths' detection.
The suggested hybrid sensing method, used to improve sensing performance, is validated using
traditional techniques but it suffers from computational complexity.
Tables [1, 2] show a summary of the previous related works of two stage SS methods [18-26].
This table provides a thorough comparison of the related two stages prior research in terms of
SNR for PU signals (SNRp), sensing time)ms), PF, and number of sensing samples N. the tables
provide a complete comparison of the earlier efforts. According to prior research, it has been
shown that the various two-stage SS approaches have a tendency to have a high computing
complexity, high sensing overheads, low and medium sensing accuracy, and bad performance in
uncertain noise environments. Finally, we observe that in the situation of low SNR values, the
step of distinguishing between PU signals and noise is missing. The ED technique alone cannot
be utilized to make decisions about the main user (PU) condition when the SNR is low. So, the
IDD detector step is utilized before to the ED to minimize noise and identify the PU signal in the
presence of noise. As a result, we offer an alternative implementation of the two stage detection
techniques which can help in substantial savings in the sensing time required to compute an
accurate result while maintaining the target detection and false alarm probabilities of the CR
device.

75
Table 1. Comparison of the related works on two-stage SS methods
Table 2. Comparison of the related works on two-stage spectrum sensing methods
N
𝐍𝐒
𝐏𝐅
sensing
time)ms)
PU
SNR
(dB)
Two Stage
Spectrum
Sensing
Techniques
N
𝐍𝐒
𝐏𝐅
sensing
time)ms)
PU
SNR
(dB)
Two Stage
Spectrum
Sensing
Techniques
1000
-
0.1
52.5
-10
ED + CD
[19]
-
-
0.1
150
-10
ED + CD
[18]
1000
-
0.1
52.8
-12
-
-
0.1
152
-12
1000
-
0.1
53
-14
-
-
0.1
153
-14
1000
-
0.1
53
-16
-
-
0.1
162
-16
1000
-
0.1
53.2
-18
-
-
0.1
178
-18
1000
-
0.1
53.2
-20
-
-
0.1
192
-20
_
_
0.1
12
-10
ED +
one order CD
[20]
1000
-
0.1
47.8
-10
MED-FT +
ED- ADT
[19]
_
_
0.1
12
-12
1000
-
0.1
48.2
-12
_
_
0.1
12
-14
1000
-
0.1
48.4
-14
_
_
0.1
12
-16
1000
-
0.1
48.4
-16
_
_
0.1
12
-18
1000
-
0.1
48.7
-18
_
_
0.1
12
-20
1000
-
0.1
48.9
-20
80000
-
0.1
70
-10
ED + CAV
modified
[21]
80000
-
0.1
78
-10
ED + CAV
general case
[21]
80000
-
0.1
70
-12
80000
-
0.1
76
-12
80000
-
0.1
70
-14
80000
-
0.1
72
-14
80000
-
0.1
68
-16
80000
-
0.1
74
-16
80000
-
0.1
68
-18
80000
-
0.1
73
-18
80000
-
0.1
70
-20
80000
-
0.1
74
-20
N
𝐍𝐒
𝐏𝐅
sensing
time)ms)
PU
SNR
(dB)
Two Stage
Spectrum
Sensing
Techniques
N
𝐍𝐒
𝐏𝐅
sensing
time)ms)
PU SNR
(dB)
Two Stage
Spectrum
Sensing
Techniques
-
-
-
128
-10
ED + CD
[23]
1000
-
0.1
47.4
-10
ED + ED-ADT
(ESNR_ADT)
[22]
-
-
-
128
-12
1000
-
0.1
47.4
-12
-
-
-
128
-14
1000
-
0.1
47.6
-14
-
-
-
200
-16
1000
-
0.1
47.6
-16
-
-
-
200
-18
1000
-
0.1
47.8
-18
-
-
-
200
-20
1000
-
0.1
48
-20

76
3. THE PROPOSED MODEL
Noise impacts the received signal randomly and can be produced by a variety of factors,
including time-varying thermal noise at the communication system's receiver and noise
interference from the surrounding environment. Furthermore, a variety of factors, such as
electromagnetic interference, affect the received signal. So noise reduction is one of the most
challenging issues in communication systems, and it has gained much attention recently. Noise
cancellation methods and technologies have evolved in response to the requirement for reducing
unwanted noise in desired signals. Wavelet transform (WT) has been shown in high-performance
signal processing tool [27-29] to be a particularly efficient technique for de-noising signals when
compared to conventional approaches such as the Fourier filter and Savitzky-Golay filter [30].
Wavelet de-noising With WT, the signal's time and frequency characteristics are presented, and
the resulting approximation and detail coefficients reflect the low and high frequency
components of the signal under analysis, respectively. As noise is mainly composed of high
frequency components, noise reduction is accomplished by thresholding the detail coefficients.
The de-noising process typically consists of the following three steps:
 Decomposition: the noisy signal is converted into a collection of orthonormal wavelet basis
functions, selecting a mother wavelet and a maximum decomposition level followed by
computing the decomposition coefficients at each level.
 Thresholding: In this stage, the threshold values for each level are evaluated, and the
coefficients at each level are applied to the threshold.
 Reconstruct: Using the inverse wavelet transform, the signal is rebuilt using the modified
coefficients.
Hence the packet of wavelet technique is a generalization of wavelet decomposition that provides
a wide variety of signal analysis scales. In packet wavelets analysis, the details as well as the
approximations are divided into 𝟐𝒏
different ways to represent the signal where the
8192
_
0.1
8
-10
ED +CFD
[25]
1000
-
0.1
45.55
-10
Improved
TSD
[(ED_SAT) +
(ED_TAT)]
[24]
8192
_
0.1
8
-12
1000
-
0.1
45.7
-12
8192
_
0.1
8
-14
1000
-
0.1
46
-16
8192
_
0.1
23.8
-16
1000
-
0.1
46.1
-16
_
_
_
_
-18
1000
-
0.1
46.3
-18
_
_
_
_
-20
1000
-
0.1
46.5
-20
1000
_
0.1
850
-10
multi-path
hybrid sensing
model
[26]
1000
_
0.1
860
-12
1000
_
0.1
850
-14
1000
_
0.1
860
-16
1000
_
0.1
870
-18
1000
_
0.1
880
-20

77
decomposition level is n. Several bases are produced by single wavelet packet decomposition,
which provides a more complicated and adaptable analysis. The best level of signal
decomposition is chosen using an entropy-based criterion. Fig. 1 displays the packet of wavelet
decomposition and reconstruction trees.
The input signal 𝑅 is divided by using filters 𝐺𝑜 and 𝐺1 into a low pass component 𝑅𝑜 and a high
pass component 𝑅1, which are decimated (down-sampled) by factor two. The low pass
component is then divided more into 𝑅00 and 𝑅01, which are again decimated by factor two. This
procedure is repeated as needed. The DWT outputs are the band pass coefficients 𝑅1 ,𝑅01,
𝑅001,……, and the last low pass coefficients 𝑅000….0 The decimation ensures that the input
sample rate and total output sample rate are identical, so there is no redundancy in the
transform. The signal is reconstructed using a pair of reconstruction filters 𝐻𝑜 and 𝐻1 are
utilized in the configuration of Fig. 1 where 𝑅000 may be reconstruction from 𝑅0000 and 𝑅0001;
and then 𝑅00 from 𝑅000 and 𝑅001; and so on back to 𝑅, using an inverse tree of 𝐻 filters.
Figure 1. packet of wavelet decomposition and reconstruction trees
When a noisy signal is de-noised using the WT approach, the accuracy of the reconstructed signal
is determined by both the thresholding method and the threshold values. There are two
fundamental methods for compressing wavelet coefficients: hard thresholding and soft
thresholding [31, 32]. Several methods for selecting thresholds have been studied [33]. Standard
techniques contain level-dependent and interval-dependent threshold selections. The level
thresholding method, which selects a single universal threshold for each level then applies it to
threshold all coefficients at the selected level. To threshold the coefficients within each interval,
the interval thresholding approach first divides the coefficients at each level into blocks and then
chooses a local threshold for each interval. This signal division makes the heteroscedastic noise
be approximately homoscedastic noise within each interval, enhancing de-noising efficiency, So
the proposed two stage model using IDD and ED for detecting the PU signals, especially in the
case of low received PU Signal to Noise Ratio (SNRp), is used to improve the sensing
performance as demonstrated in the simulation result below because it obtains a higher PD and a
lower PF , even at low SNRs and gains minimum sensing time due to the simplicity and less
complexity of the proposed method. Flowchart of the proposed model is shown in Fig. 2. The
IDD stage is used for noise reduction and for extraction of the PU signals from noise at low PU
SNR to perform reliable detection and accurate SS.

78
Figure 2. Flowchart of the proposed model
The ED detector is performed in the second stage due to its efficiency and simplicity. To compute
the energy of the received signal, the samples are squared and integrated during the observation
period; the integrator output is then compared to the threshold. If the integrator's output higher
than the threshold, it will be assumed that the given radio spectrum is occupied. Otherwise, it is
considered to be empty. The following hypothesis for received signal [34] can be applied to
identify the PU signal
S(n) =
w(n)HP = H0
g(n) ∗ 𝑥𝑝(n) + w(n) HP = H1
(1)
where the received signal is represented by 𝑆(𝑛) and the channel gain is represented by g(𝑛), the
PUsignal of a transmitted power 𝑃𝑝 is represented by 𝑥𝑝(𝑛) , the additive white Gaussian Noise
(AWGN) at the cognitive user receiver is represented by 𝑤(𝑛), which is considered to be a
Circularly Symmetric Complex Gaussian (CSCG) random process with zero mean and one-sided
power spectral density 𝑁𝑜 (i.e., 𝑤(𝑛)~ 𝒞𝒩(0, 𝑁𝑜)), and 𝑛 represents the time index. The null
hypothesis is represented by 𝐻0, which represents the absence of PU and the alternative
hypothesis is represented by 𝐻1, which represents that PU is present.
For ED, the decision factor 𝐷f𝐸𝐷 is computed by the energy of the samples that were taken during
the observation window duration τ, and it is expressed by:
𝐷𝑓𝐸𝐷 =
1
𝑁
∑|S(n)|2
𝑁
𝑛=1
(2)
Where N represents the number of sensing samples (𝑁 = 𝜏𝑓𝑠) with 𝑓𝑠 indicating the sampling
frequency
By comparing the decision factor with a predefined threshold 𝜓 the probabilities of false alarm
and detection can be determined, which determines under the both hypothesis 𝐻𝑃 = 𝐻0 and 𝐻𝑃 =
𝐻1, respectively.
𝐷𝑓𝐸𝐷
≥
<
𝜓 (3)
𝑃𝐹 And PD can be determined as follows:

79
𝑃𝐹 = 𝑃𝑟𝑜𝑏( 𝐷f𝐸𝐷 > 𝜓 |𝐻𝑃 = 𝐻0 ) = 𝑄(
𝜓−𝜇0
𝜎0
) (4)
PD = 𝑃𝑟𝑜𝑏( 𝐷f𝐸𝐷 > 𝜓 |𝐻𝑃 = 𝐻1 ) = Q(
𝜓−μ1
σ1
) (5)
Where 𝑄(𝑥 ) =
1
√2𝜋
∫ 𝑒−
𝑡
2 𝑑𝑡
∞
𝑡
denotes the complementary distribution function of the standard
Gaussian
The decision factor 𝐷f𝐸𝐷 has a normal distribution then the mean and variance have been shown
as:
μ0
= No , σ0
2
=
1
N
No
2
(6)
μ1
= (γ + 1)No, σ1
2
=
1
N
(γ + 1)2
No
2
(7)
Let 𝛾 = |g𝑝|
2
𝛾𝑃represent the SNR of the received signal
𝑃𝐹 , PD After substitution of mean and variance values:
𝑃𝐹(𝜓, 𝜏) = 𝑄 ((
𝜓
𝑁𝑜
− 1) √𝑁) = 𝑄 ((
𝜓
𝑁𝑜
− 1) √𝜏 𝑓𝑠) (8)
𝑃𝐷(𝜓, 𝜏) = 𝑄 ((
𝜓
(𝛾+1)𝑁𝑜
− 1) √𝜏 𝑓𝑠) (9)
To ensure that a PU network is effectively protected from harmful interference from a secondary
network, typically a limitation is adopted on the targeted 𝑃𝐷 , especially in a low SNR
environment. In IEEE 802.22 WRAN, the targeted probability of detection 𝑃𝐷 = 0.9 for the SNR
of −20 dB
For the target𝑒𝑑 𝑃𝐷 the decision threshold can be obtained from Equation (9) as follows:
𝜓 = (𝛾 + 1)𝑁𝑜 (
1
√𝜏 𝑓𝑠
𝑄−1(𝑃𝑑 ) + 1) (10)
It can rewrite them such that one is a function of the other. The detection probability is related to
the false alarm probability as follows:
𝑃𝐷 = 𝑄 (
1
(𝛾 + 1)
(𝑄−1
(𝑃𝑓) − 𝛾√𝜏 𝑓𝑠)) (11)
𝑃𝐹(𝜏) = 𝑄 ((𝛾 + 1)𝑄−1(𝑃𝑑 ) + 𝛾√𝜏 𝑓𝑠) (12)
The IDD technique is implemented in the first stage for noise reduction and efficient SS process,
in signal de-noising applications Wavelet transformation is widely employed the sampled signal
inside the band of the licensed user must always be processed by the CR. the system equations in
(1) are simplified as follows:
S1 = w(n) (13)
S2 = 𝑥𝑝(n) + w(n) (14)

80
The wavelet transform of the received signal is shown as follows:
[𝑎𝑆, 𝑑𝑆] = 𝑊𝑆 = W(𝑥𝑝 + 𝑤) = 𝑊
𝑥𝑝
+ 𝑊𝑊 (15)
Where W is the left invertible transformation matrix of the DWT, because 𝑥𝑝 contains few
changes the detail information of 𝑊
𝑥𝑝
is sort of zero. Therefore, the detail information
dS contains the details that indicate the majority of the noise power in the wavelet transformation
domain. After thresholding the detail information, the inverse wavelet transform can recover the
desired signal with less noise effect which increase the ability to distinguish between the signal
and noise which improving the accuracy of ED technique. Finally, the proposed method assists in
overcome the problems of high Pf rates, poor detection accuracy at low SNR levels, and noise
uncertainties and helps to gain minimum sensing time which required in CR networks.
4. SIMULATION RESULTS FOR THE PROPOSED MODEL
In the presented system model, the simulation results and analysis of the proposed model are
executed in MATLAB, using the sampling frequency of 6 MHz, a zero-mean additive white
Gaussian Noise and one-sided power spectral density 𝑁𝑜 = 1 over 1000 Monte Carlo simulations,
at low SNRs most previous techniques are not robust to noise, therefore impacting CR's
performance and reliability. The simulation is performed at an SNR varying from −10 to -20 dB
which is the most common range for low SNR in previous related two stage methods and N
varying from 1000 to 100000, P (𝐻1) = 0.2 and the targeted PD = 0.9 according to IEEE 802.22
WRAN. The time needed by CR user to detect the licensed PU signal is the SS time. The PU can
make more efficient use of its band if the sensing time is increased and the limit is set so that CR
user is unable to interfere with PU at that long time. The longer spectrum sensing, the more PUs
will be detected and the less interference will occur since PUs can make maximal utilization of
their priority right. The sensing time is directly proportional to the samples number received by
the CR user. Table 3 contains the result of PF of the proposed model at different signal to noise
ratio at N equal 1000, 10000,100000 respectively.
Table 3. results of proposed model
N
𝐍𝐒
𝐏𝐅
sensing
time)ms)
PU
SNR
(dB)
Two Stage
Spectrum
Sensing
Techniques
N
𝐍𝐒
𝐏𝐅
sensing
time)ms)
PU
SNR
(dB)
Two Stage
Spectrum
Sensing
Techniques
10000
1000
0
1.66
-10
Proposed
model
1000
1000
0.03
0.3
-10
Proposed
model
10000
1000
0
1.66
-12
1000
1000
0.28
0.3
-12
10000
1000
0
1.66
-14
1000
1000
0.52
0.3
-14
10000
1000
0.12
1.66
-16
1000
1000
0.71
0.3
-16
10000
1000
0.39
1.66
-18
1000
1000
0.78
0.3
-18
10000
1000
0.61
1.66
-20
1000
1000
0.85
0.3
-20
100000
1000
0
16.66
-10
Proposed
model
100000
1000
0
16.66
-12
100000
1000
0
16.66
-14
100000
1000
0
16.66
-16
100000
1000
0
16.66
-18
100000
1000
0.03
16.66
-20

81
Figure 3. Probability of false alarm versus signal to noise ratio for different values of N
As shown in figure 3 there is an inverse relation between the value of PF and the signal to noise
ratio, when N = 1000 then PF = 0.03 at SNR = −10 dB and PF = 0.85 at SNR = −20 dB. When N
= 10,000 then PF = 0 at SNR = −10 dB and PF = 0.61 at SNR = −20 dB. When N = 40,000 then
PF = 0 at SNR = −10 dB and PF = 0.24 at SNR = −20 dB. When N = 100,000 then PF = 0 at
SNR = −10 dB and PF = 0.03 at SNR = −20 dB.
Figure 4. shows the graph of spectrum sensing time versus Number of samples from 1000 to 10000.
-20 -19 -18 -17 -16 -15 -14 -13 -12 -11 -10
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
SNR
Probability
of
false
alarm
N = 1000
N = 10000
N = 40000
N = 100000
0 0.5 1 1.5
1000
2000
3000
4000
5000
6000
7000
8000
9000
Number
of
samples
Sensing time(ms)

82
Figure 5. Shows the graph of spectrum sensing time versus Number of samples from 10000 to 100000.
As targeted in IEEE 802.22 standards that the time taken by CR users during the detection of PU
spectrum bands for SS should be as small as possible. Figures 4, 5 show graphs of mean detection
time of the proposed model versus various values of samples. As shown it is clear that the
detection time increases with the increase of samples and these results are arranged at table 4
which shows that the proposed model take better time than other two stage techniques in table
[1,2] .
Table 4. the proposed model sensing time for different values of N
N
𝐍𝐒
Sensing time
(ms)
Two Stage Spectrum
Sensing Techniques
1000
1000
0.3
Proposed model
2000
1000
0.35
4000
1000
0.65
6000
1000
1
8000
1000
1.38
10000
1000
1.66
20000
1000
3.4
40000
1000
6.8
60000
1000
10
80000
1000
13.4
100000
1000
16.6
0 2 4 6 8 10 12 14 16 18
1
2
3
4
5
6
7
8
9
10
x 10
4
Number
of
samples
Sensing time(ms)

83
Figure 6. shows the graph of Sensing time versus SNR for different values of N
Finally, figure 6 shows the graph of Sensing time versus SNR for different values of N, From the
Figure it is clear that the detection time increases with the value of N but the detection time will
be less than the other two stage techniques in table [1, 2].
5. CONCLUSION
An improved spectrum sensing model and its performance evaluation are introduced in this
paper. In this paper, a new model is discussed using IDD and ED technique to improve the
different spectrum sensing parameters. The interval dependent de-noising technique is applied in
the first stage to improve the reconstruction of the required signal in order to reduce the noise
effect and to detect the PU signals accurately which make the proposed model achieves the
desired target PD and reduces PF especially at low SNR as well as reduced sensing time compared
to previous two stage sensing detector, The Simulation results of this method are utilized to
demonstrate the reduction in sensing time. The obtained results demonstrated that the proposed
model using IDD-ED is very better in terms of sensing time and false alarm. Different SNR
values are taken to calculate the sensing time and probability of false alarm. The simulation
results declared that the proposed model achieved alow probability of false alarm and minimum
sensing time.
STATEMENTS AND DECLARATIONS
• Funding: no funding was received to assist with the preparation of this manuscript.
• Financial interests: The authors declare they have no financial interests.
The authors have no competing interests to declare that are relevant to the content of this article.
• Consent to participate: not applicable
CONFLICTS OF INTEREST
The authors declare no conflict of interest.
-20 -19 -18 -17 -16 -15 -14 -13 -12 -11 -10
0
2
4
6
8
10
12
14
16
18
SNR
sensing
time(ms)
N=1000
N=10000
N=40000
N=100000

84
REFERENCES
[1] FCC, Second report and order and memorandum opinion and order, ET Docket No. 08-260, 2008.
[2] Li, C., 2009.Department of Electronic Engineering, School of Electronic, Information and Electric
Engineering,Shanghai Jiao Tong University, Report.Retrievedfrom:http://iwct.sjtu.edu.cn/
Personal/xwang8/research/CR_zhu/project.html.
[3] Khobragade, A. S., & Raut, R. D. (2017). Hybrid spectrum Sensing method for cognitive radio.
International Journal of Electrical and Computer Engineering (IJECE), 7(5), 2683.
https://doi.org/10.11591/ijece.v7i5.pp2683-2695.
[4] Yue, W. and B. Zheng, 2009. A two-stage spectrum sensing technique in cognitive radio systems
based on combining energy detection and one-order cyclostationary feature detection. Proceeding of
the International Symposium on Web Information Systems and Applications (WISA’09). Nanchang,
P.R. China, pp: 327-330.
[5] I. F. Akyildiz, W.-Y. Lee, M. C. Vuran and S. Mohanty, Next generation/dynamic spectrum
access/cognitive radio wireless networks: A survey, Computer Networks, vol.50, no.13, pp.2127-
2159, 2006.
[6] J. Ma, G. Li and B. H. Juang, Signal processing in cognitive radio, Proc. of the IEEE, vol.97, no.5,
pp.805-823, 2009.
[7] T. Yucek and H. Arslan, A survey of spectrum sensing algorithms for cognitive radio applications,
IEEE Communications Surveys Tutorials, vol.11, no.1, pp.116-130, 2009.
[8] S. Haykin, Cognitive radio: Brain-empowered wireless communications, IEEE Journal on Selected
Areas in Communications, vol.23, no.2, pp.201-220, 2005.
[9] A. Fehske, J. D. Gaeddert and J. H. Reed, A new approach to signal classification using spectral
correlation and neural networks, Proc. of IEEE DySPAN, pp.144-150, 2005.
[10] Z. Liu and J. Wang, A Hermitan mapping approach to unitary root-music algorithm for
cyclostationary signals, International Journal of Innovative Computing, Information and Control,
vol.5, no.10(A), pp.3123-3132, 2009.
[11] W. Yue and B. Zheng, Spectrum sensing algorithms for primary detection based on reliability in
cognitive radio systems, Journal of Compute and Electrical Engineering, vol.36, no.3, pp.469-479,
2010.
[12] W. Yue, B. Zheng, Q. Meng and W. Yue, Combined energy detection and one-order cyclostationary
feature detection techniques in cognitive radio systems, The Journal of China Universities of Posts
and Telecommunications, vol.17, no.4, pp.18-25, 2010.
[13] I. F. Akildiz, B. F. Lo and R. Balakrishan, Cooperative spectrum sensing in cognitive radio
networks: A survey, Physical Communication, vol.4, pp.40-62, 2011.
[14] I. W. G. on Wireless Regional Area Networks (WRANs), http://www.ieee802.org/22/.
[15] C. R. Stevenson, C. Cordeiro, E. Sofer and G. Chouinard, Functional requirements for the 802.22
WRAN standard, IEEE 802.22 Draft, 2005.
[16] K. Srisomboon, K. Thakulsukanant, A. Prayote, and W. Lee, “Adaptive two-stage spectrum sensing
under noise uncertainty in cognitive radio networks,” ECTI Transactions on Electrical Engineering,
Electronics, and Communications, vol. 14, no. 1, pp. 21-35, 2016.
[17] A. Ghasemi and E. S. Sousa, Spectrum sensing in cognitive radio networks: Requirements,
challenges and design trade-offs, IEEE Communication Magazine, pp.32-39, 2008.
[18] Maleki, S., Pandharipande, A., & Leus, G. (2010, March). Two-stage spectrum sensing for
cognitive radios. In 2010 IEEE International Conference on Acoustics, Speech and Signal
Processing (pp. 2946-2949). IEEE.
[19] Bagwari, A., & Tomar, G. S. (2013). Two-stage detectors with multiple energy detectors and
adaptive double threshold in cognitive radio networks. International Journal of Distributed Sensor
Networks, 9(8), 656495.
[20] Ejaz, W., Hasan, N. U., & Kim, H. S. (2012). SNR-based adaptive spectrum sensing for cognitive
radio networks. Int. J. Innov. Comput. Inf. Control, 8(9), 6095-6106.
[21] Geethu, S., & Narayanan, G. L. (2012). A novel high speed two stage detector for spectrum sensing.
Procedia Technology, 6, 682-689.
[22] Bagwari, A., Kanti, J., Tomar, G. S., & Samara, A. (2016). A robust detector using snr with
adaptive threshold scheme in cognitive radio networks. International Journal of Signal Processing,
Image Processing and Pattern Recognition, 9(5), 173-186.

85
[23] Minny Bhola , Rinkoo Bhatia , Shruti Tiwari ,"Two Stage Spectrum Sensing for Cognitive Radio
using Cyclostationarity detection and Energy Detection" International Journal of Latest Trends in
Engineering and Technology (IJLTET) Vol. 1 Issue 4 November 2012.
[24] Kanti, J., & Tomar, G. S. (2018). Solution of sensing failure problem: an improved two-stage
detector. The Computer Journal, 61(6), 847-855.
[25] Nair, P. R., Vinod, A. P., & Krishna, A. K. (2011, September). A fast two stage detector for
spectrum sensing in cognitive radios. In 2011 IEEE.
[26] Rabie Mohamed, A., A. Aziz El-Banna, A., & A. Mansour, H. (2021). Multi-path hybrid spectrum
sensing in cognitive radio. Arabian Journal for Science and Engineering, 46, 9377–9384.
[27] Rioul, O., & Vetterli, M. (1991). Wavelets and signal processing. IEEE signal processing magazine,
8(4), 14-38.
[28] Mallat, S. G. (1989). A theory for multiresolution signal decomposition: the wavelet representation.
IEEE transactions on pattern analysis and machine intelligence, 11(7), 674-693.
[29] Teolis, Anthony, and John J. Benedetto. Computational signal processing with wavelets. Vol. 182.
Boston, MA, USA: Birkhäuser, 1998.
[30] Mittermayr, C. R., Nikolov, S. G., Hutter, H., & Grasserbauer, M. (1996). Wavelet denoising of
Gaussian peaks: a comparative study. Chemometrics and Intelligent Laboratory Systems, 34(2),
187-202.
[31] Donoho, D. L. (1995). De-noising by soft-thresholding. IEEE transactions on information theory,
41(3), 613-627.
[32] Alsberg, B. K., Woodward, A. M., & Kell, D. B. (1997). An introduction to wavelet transforms for
chemometricians: A time-frequency approach. Chemometrics and intelligent laboratory systems,
37(2), 215-239.
[33] Walczak, Beata, ed. Wavelets in chemistry. Elsevier, 2000.
[34] Ghasemi, A., & Sousa, E. S. (2005, November). Collaborative spectrum sensing for opportunistic
access in fading environments. In First IEEE International Symposium on New Frontiers in
Dynamic Spectrum Access Networks, 2005. DySPAN 2005. (pp. 131-136). IEEE.

Sensing Time Improvement Using Two Stage Detectors for Cognitive Radio System

More Related Content

Similar to Sensing Time Improvement Using Two Stage Detectors for Cognitive Radio System (20)

More from IJCNCJournal (20)

Recently uploaded (20)

Sensing Time Improvement Using Two Stage Detectors for Cognitive Radio System