On the Performance Analysis of Blind Spectrum Sensing Methods for Different Communication Channels

International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 04 Issue: 11 | Nov -2017 www.irjet.net p-ISSN: 2395-0072
© 2017, IRJET | Impact Factor value: 6.171 | ISO 9001:2008 Certified Journal | Page 1203
On the Performance Analysis of Blind Spectrum Sensing Methods for
Different Communication Channels
Cebrail ÇİFLİKLİ1, Fatih Yavuz ILGIN2
1Professor, Department of Electronics and Automation, Erciyes University Kayseri Vocational College,
Melikgazi, Kayseri, Turkey
2Instructor, Department of Electronics and Automation, Erzincan University Vocational College,
Erzincan, Turkey
---------------------------------------------------------------------***---------------------------------------------------------------------
Abstract - Lately, with the increase of service standard in
wireless communication systems, problem of spectrum
shortage has appeared. In order to overcome this problem,
it is necessary to use the existing frequency spectrum in the
most efficient way. Cognitive radio systems are the
emerging technologies to solve these problems. The first step
in cognitive radio systems is the detection of the full / empty
state of the current spectrum. Blind methods for this
detection are highly preferred in terms of ease of
implementation and computational cost. In this study,
performance analyzes of Rayleigh, nakagami-m, nakagami-
n and nakagami-q fading communication channels of blind
spectrum detection methods are performed. A MIMO-OFDM
based communication system was used in the study.
Simulations were performed in MATLAB environment.
Key Words: Detection theory, Spectrum sensing, Tracy-
Widom distribution, Wireless communication channel.
1.INTRODUCTION
The frequency spectrum required for wireless
communication systems is inherently limited and highly
valuable. In many countries, the available spectrum is
almost entirely allocated and the problem of spectrum
shortage has arisen [1,2]. The main factor that causes
spectrum scarcity is inefficient use of the spectrum due to
static and inflexible assignment. In the current wireless
networks, the fixed spectrum access (FSA) policy is
followed to support many different applications and
services in a way that does not interfere with each other.
This principle divides the current radio spectrum into
frequency bands dedicated to different services such as
mobile, fixed services, satellite services, radio and
television broadcasts. A certain bandwidth has been
allocated for each service and has been allocated for a long
time to license holders. Thus, while only licensed users can
use the assigned band, they are not allowed to use the
band by other users, even if the band is empty [3,4].
With the new generation of technologies, while there are
frequency bands in certain spectrum bands, the rate of use
in most of the spectrum is very low. For this reason,
spectrum efficiency is low. Thus, the efficient use of the
frequency spectrum, which is a limited source, has become
even more important. Along with the predicted increase in
data amount per user and average data rate, especially
between 2020 and 2040, the most fundamental problem
that next generation wireless technologies will encounter
is to find a free frequency band that can meet this
projected demand [1,2,5,6]. Dynamic spectrum access
techniques have been developed to provide this probing
solution, facilitate access to the spectrum, exploit more
users and achieve the maximum possible throughput.
Cognitive radio (Cognitive Radio, BR), which allows these
techniques, is emerging as one of the most promising and
most prominent next generation wireless technologies,
because it can increase productivity and increase the
number of users and service demands with a solution to
spectrum inadequacy [3,4] .
Cognitive radio is defined as an intelligent wireless
communication system capable of dynamically changing
transmitter parameters (such as frequency band,
transmitting power, modulation type) in order to prevent
interference and improve transmission speed, which is
aware of the radio environment and depends on its
interaction with the environment. The concept of cognitive
radio was first proposed by Joseph Mitola and Gerald Q.
Maguire in 1999, and as a result, the IEEE 802.22 standard
has been developed targeting the use of cognitive radios in
Wireless Regional Area Networks (WRAN)[[1-4].
The basic functions of the cognitive radios are spectrum
sensing, spectrum management (spectrum decision),
spectrum sharing and spectrum mobility. In this study,
one of the most important functions of cognitive radios,
the methods used for spectrum sensing, has been
investigated under different communication channels. The
purpose of spectral censoring is to detect the mobility of
the licensed users and the state of the spectrum by
perceiving the spectrum periodically. Many different
methods have been proposed in the literature such as
energy detection, filter equalization, such as cyclic
stationary feature detection detection, eigenvalue

detection for spectral detection. Some of the eigenvalue
based methods and some of them are defined as blind sensing.
That is, they do not require prior knowledge of the
signals. Each of these methods has certain requirements,
advantages and disadvantages. Energy sensing is one of
the most common spectrum sensing methods because it
requires no knowledge of the primary user signal in
addition to its low mathematical and hardware complexity
[5,6]. However, it is a weakness of this method that the
difficulty in determining the threshold value (the noise
variance is not known correctly), the ability to distinguish
interference from the primary user from the primary user
while using the secondary user band, and the reduction in
signal to noise ratio (SNR) values. Various feature
detectors (such as cyclic stationary feature detection) have
been proposed to improve the detection performance by
using known features of the signals to be detected and to
prevent the problem that the noise variance is not known
correctly due to the loss of performance in case of noise
uncertainty in the energy detection method [7]. Among the
methods used in spectral censoring, the blind-eigenvalue
based detection method [8-12] is predominant because it
requires no information about noise variance and source
signal. Due to multi-path damping and shadowing effects,
the problem of performance drop in spectral estimation
methods is significantly reduced by the use of multiple
antennas in the cognitive radio receiver.
Different methods have been proposed for blind-
eigenvalue detection in the literature. Detectors that
process based on the maximum-minimum eigenvalue ratio
(MME), the maximum eigenvalue-to trace (MET) ratio, and
the maximum eigenvalue-eigenvalue sum ratio are the
most commonly used for eigenvalue detection [11,12].
The aim of this study is to obtain performance analyzes of
some blind-eigenvalue based detection methods in the
literature with different fading communication channels.
In this work, it specifies the matrices of the bold lowercase
(x) letters, and the normal lowercase (x) letters indicate
the vectors. (x '), the variable specifies the transpose.
2. BLIND EIGENVALUE BASED SENSING AND
WIRELESS COMMUNICATION CHANNEL
The main purpose in spectrum sensing is to detect the
presence of a primary user signal in only a specific region
of the frequency spectrum. For multi-antenna systems, the
following scenario can generally be used for spectrum
detection. Where the secondary users may be randomly
distributed. The prominent point is that they are within
the coverage area of the primary user of the secondary
user. The redundant number of antennas in the secondary
users increases the detection performance. but increasing
the number of antennas can be difficult in practice.
Fig-1: Proposed scenario for eigenvalue based spectrum
sensing
Secondary users equipped with multiple antennas should
determine that the spectrum is full / empty. When the
spectrum is empty, they use the relevant spectrum to
communicate among themselves. However, when the
primary user becomes active, they must immediately
release the relevant spectrum. Where h1, h2 hp represent
the channel coefficient vector from the primary user to the
secondary user [13]. Detection problem is mathematically
defined as follows.
(1)
(2)
Where, x represents the matrix of the secondary user's
signal through the multiple antennas, s is the gaussian and
zero mean primary user signal, and n is the gaussian noise.
Therefore, in case the primary user is inactive, the matrix x
only consists of noise. However, in the case of H1, the
matrix x recognizes both the primary user mark and the
noise signals mixture. Thus the eigenvalue distributions of
the covariance matrix of the matrix x for the cases H1 and
H0 are as follows.
(3)
(4)
Where p shows the number of antennas in secondary users.
There will be a difference in the probability distribution
functions and variances of the eigenvalues according to the
active / passive state of the primary user as seen in the
equations. For the eigenvalue based detection, a decision is
made between the states H1 and H0 using this difference.
Equations 3 and 4 'represent the channel coefficient vector.
There are some channel simulations for digital
communication systems. The major ones are shown below.
Birincil verici
TX
İkincil Kullanıcı
RX
İkincil Kullanıcı
RX
h1
hm
h2

2.1. Rayleigh fading channel
The Rayleigh distribution is often used to model multi-
path damping (NLOS - Non line of sight) situations (such
as mobile systems). Rayleigh fading is expressed by the
following probability density function (PDF-Probability
Density Function), with the channel amplitude fading [13]:
( ) (5)
Where, the mean-square value of the fading amplitude is
given as . In this fading model, the signals are
independent of each other and reflect to the receiver.
2.2. Nakagami-n fading channel
Nakagami-n distribution is also known as the Rice
distribution. It is used for modeling propagation paths
consisting of a strong line of sight (LOS) component and a
large number of randomly weak components (Figure 2).
The fading amplitude of the channel is defined by the
following PDF [13].
( )
(
( )
) (
√
) (6)
Where n is the fading parameter that can take values in the
range 0 to ∞, and I0 is the first type of zero-order modified
Bessel function. The Nakagami-n distribution represents
rayleigh fading in the case n= 0 and non-damping states in
the case of n =∞.
Fig-2: LOS-Line of Sight and NLOS- Non Line of Sight
Channel
2.3. Nakagami-q(hoyt) fading channel
The Nakagami-q distribution is observed in radio waves
that are strongly bound to strong ionospheric fading. The
PDF for the Hoyt distribution is defined as [13].
( )
(
( )
) (
( )
) (7)
where q is a fading parameter that can take values in the
range 0 to 1.
2.4. Nakagami-m fading channel
Nakagami-m fading distribution is commonly used by
Nakagami to model fading channels. The Nakagami-m
distribution has the center of the chi-square as follows
[13].
( ) (8)
where m is the fading parameter and Γ (m) is the Gamma
function and is defined as ∫ . As the
value of the fading parameter increases, the damping
intensity decreases.
2.5. Threshold for MME
A decision between the H1 and H0 states in the narrow
band spectrum detection methods depends on the test
statistic (TS) and the threshold( ) value. This is
mathematically defined as follows [12,15,16];
(9)
(10)
Where indicates false alarm probability and indicates
probability of detection. When calculating the threshold value,
there is a limit value of 0.1, as determined by the 802.11
wireless communication working group, and this value should
be considered. The threshold value should be set to , ie the
probability of false detection. γ is the threshold value.
(11)
Where denotes the largest eigenvalue of the covariance
matrix of x, and denotes the smallest eigenvalue.
If the equality is rearranged because the H0 hypothesis is
valid, the following equation is obtained as follows;
(12)

One side of the equation should be imitated to the Tracy-
widom distribution of order 1. Where (√ √ ) should be
written in place of . In this case, the following equation
is obtained. Where n is the number of samples, p is the
number of antennas in the secondary user.
(√ √ ) (13)
Where one side of the equation should be likened to tracy
widom distribution of erder 1. Tracy-widom distribution of
order 1 is described below [17].
→ (14)
Where some parameters defines are as follows.
(√ √ ) (15)
(√ √ ) (
√ √
)
⁄
(16)
(17)
shows the signal samples received under the H0
hypothesis, that is, when there is only noise in the channel. In
order to compare Equation 13 to tracy-widom distribution of
order 1, the following additions have to be made.
(( ) (
√ √
)) (18)
If survival function is used, the threshold value for MME is
obtained as follows.
( ) (
(√ √ )
(√ √ )
)
(
(√ √ )
⁄
⁄ )(19)
Where denotes the tracy-widom distribution of order 1.
this distribution is the probability distribution of the largest
eigenvalue of the random hermit matrix. The specific values
for this distribution are given in Table 1.
Tablo -1: Numerical table for tracy widom distribution of
order 1.
x -3.90 -2.78 -1.27 0.45 2.02
F1(x) 0.01 0.10 0.50 0.90 0.99
2.6.Threshold for MET
There is a test statistic in the MET method as the ratio of the
largest eigenvalue of the covariance matrix to the sign of the
received signal.
(20)
Where specifies trace of the received signal. In this
method, the following equation is obtained if the equation is
rearranged to obtain the threshold value.
(21)
Here equation 21 must be likened to the Tracy-widom
distribution of order 1. Thus the threshold value for the MET
method is defined as follows.
[ ( ) ⁄ ]
(22)
Where α and β denote the bevel and variance coefficients for
the Tracy-widom distribution and are defined as follows [9].
(√ √ ) (23)
√ √ (
√
)
√
(24)
3. SIMULATION
In this section, simulation results are given to see the
success of the proposed methods. MIMO-OFDM based
communication system is used in simulations. Figure 3
shows the detection performance of the proposed
methods against varying SNR values for a 3x3 MIMO
system (3 receivers, 3 transmit antennas). Pfa = 0.1 (this is
the limit value allowed by WRAN 802.22 working group).
In the simulations, the primary user signal and noise
signal were randomly generated, and each algorithm was
run 1000 times for monte carlo analysis to obtain the
average of the detection probability values. Referring to
Fig. 3, it is seen that the detection performance is the most
successful for the Rayleigh fading channel for the MME
method. Nakagami-m and nakagami-n damped channels
show very close performance to each other.
The nakagami-n is used to model propagation paths
consisting of a fading channel, a strong line of sight (LOS)
component, and a large number of randomly weak
components. This usually refers to the communication
channels from the satellite to the base stations or vice
versa. On the contrary, rayleigh damping can be shown as
a communication network in the digital communication

systems of today with mobile phones and base station or
vice versa. Because there is no line of sight (LOS)
component at rayleigh fading. In the simulations, the
Rayleigh channel at the rate of -9 dB SNR for the MME
method exhibited a perfectly accurate detection. If the
noise level increases further, we see that the probability of
detection is reduced.
Fig-3: Detection performances in different communication
channels for the MME method n = 1000, p = 3.
As can be seen, the nakagami-n channel provides a
detection performance in the middle of the other three
channels. No additive white gaussian noise (AWGN)
channel simulations are included in the simulations. The
reason for this is that the AWGN channel models any faded
channels. If the simulation results for the AWGN channel
were in the graphs, it could have the best possible
perceived performance. However, because the AWGN
channel does not contain any fading, it is not the preferred
choice for simulations in digital communication systems.
Performance analysis of the MET based detection method
in different fading channels is given in Fig 4. Again, as
shown here, the best sensing performance is presented
under Rayleigh fading channel.
Fig-4: Detection performances in different communication
channels for the MET method n = 1000, p = 3.
If Figure 3 is compared with Figure 4, it is seen that MME
method perceives more successfully than MET. In
addition, MME method is more advantageous than MET in
terms of cost of calculation.
The performance of the blind spectrum detection methods
according to the sample length is shown in Fig-4. As it is
understood from the previous graphs that the most
successful method is MET, only this method is given for
the sample length. It is seen here that the probability of
detection increases with the increasing number of samples
in direct proportion. In cognitive radio systems, increasing
the sample length for the relevant frequency to be used as
an opportunist is also an unwanted condition. This is
because the increase in the sample length means that the
number of people who are in the process of perceiving the
speculum is also increasing. For this to be accurate at least
the length of the sample is an important point for cognitive
radio systems.
Fig-4: Detection performances for different sample
lengths for the MME method.
3. CONCLUSIONS
Since the spectrum is a scarce and ending source, using
this source efficiently is an important point for the future
of wireless communication systems. Over the last years,
much work has been done to solve this problem under the
topic of digital communication. Blind methods for
spectrum sensing have been highly promising since they
do not require any prior knowledge for the primary user
and the noise signal. The success of spectrum sensing
methods is greatly influenced by the dynamic
characteristics of the communication environment. The
channel modelling techniques and methods already used
for digital communication are still among the over-worked
topics. For this reason, the performance of the spectrum
sensing methods using the greatest -minimum eigenvalue
and the largest eigenvalue-ratio is investigated in this
study. In simulations, MME and MET methods offered the
best performance on rayleigh damped channel.
REFERENCES
[1] Sudharman K. J.,
.
-12 -10 -8 -6 -4 -2 0
0
0.5
1
SNR
Probabilityofdetection(Pd
)
Nakagami-m
Nakagami-n
Rayleigh
Nakagami-q
-10 -8 -6 -4 -2 0 2
0
0.5
1
SNR
)
Nakagami-m
Nakagami-n
Rayleigh
Nakagami-q
-20 -15 -10 -5 0
0
0.5
1
SNR
)
100 sample
200 sample
500 sample
1000 sample
5000 sample

[2] Cognitive Radio: making
Software Radios More. IEEE Pers. Commun.,6,
4(1999), pp. 1318.
[3] T. Ratnarajah, R. Vaillancourt ; Alvo M.; ‘’Eigenvalues
and condition num mp x m m ’’
SIAM J. Matrix Anal.Appl., 26, 2(2005), pp. 441–456.
[4] M. Mishali, Y. Eldar , ‘’Blind Multiband Signal
Reconstruction: Compressive Sensing for Analog
Signals’’ , IEEE Trans. Sig. Proc., 57, 3(2009), pp. 993–
1009.
[5] I. Sobron ,P. Diniz , W. Martins , Velez M. ‘’Energy
detection technique for adaptive spectrum sensing’’ ,
IEEE Trans. Commun. 63 (3) (2015) 617–627
[6] S. Chatterjee, A.Banerjee , Acharya T.; Maity S. P,
‘’Fuzzy c-means clustering in energy detection for
cooperative spectrum sensing in cognitive radio
system’’ in: Proceedings of Multiple Access
Communications (MACOM) Sweden, 8715, (2014), pp.
84–95.
[7] Y. Mingchuan ,L. Yuan ,L. Xiaofeng ,T. Wenyan,
‘’Cyclostationary feature detection based spectrum
sensing algorithm under complicated electromagnetic
environment in cognitive radio networks ‘’, China
Commun. 12 (9) (2015) 35–44.
[8] D. Bhargavi , C. Murthy , ‘’Performance comparison of
energy, matched-filter and cyclostationarity-based
spectrum sensing’’ Proceedings of IEEE Eleventh
International Workshop on Signal Processing
Advances in Wireless Communications (SPAWC)
,Morocco, 2010, pp. 1–5.
[9] Y. Zeng , Y.C. Liang , ‘’Covariance based signal
detections for cognitive radio’’ Proc IEEE
International Symposium on and New Frontiers in
DynamicSpectrum Access Networks (DySPAN 2007) ,
Dublin, Ireland, April 2007, 2007, pp. 202–207.
[10] M. Jin , Q. Guo ,J. Xi , Y. Youming Li ,Y. Yu , Huang D.
‘’Spectrum Sensing Using Weighted Covariance Matrix
in Rayleigh Fading Channels’’ EEE T ı O
V ı T 6 ( 5) pp 5 37-5148.
[11] N. Pillay , H.J. Xu, ‘’Blind Eigenvalue-Based Spectrum
Sensing for Cognitive Radio Networks’’. IET Comm.
6,11 (2011), pp.1388-1396.
[12] C. Çiflikli , F.Y .Ilgin; ‘’Covariance based spectrum
sensing with studentized extreme eigenvalue’’ ,
Technical gazette (accepted papers in press) DOI:
10.17559/TV-20161217120341
[13] N. Kapucu ,’’ ş kç şm m k ı
sönümlü k k ş ım z ç
matematiksel ifadelerin türetilmesi’’,Kayseri,2012
[14] S.Y.Xu , Y.L.Shang Y and H. M. Wang , ‘’Double
thresholds based cooperative spectrum sensing
against untrusted secondary users in cognitive radio
networks’’, in Proc. 2009 IEEE Vehicular Technology
Conference , vols. 1-5, pp. 519–523.
[15] C. Charan, R. Pandey, ‘’Eigenvalue Based Double
Threshold Spectrum Sensing Under Noise Uncertainty
for Cognitive Radio’’, Optik - International Journal for
Light and Electron Optics . 127,15, (2016), pp 5968–
5975.
[16] L. Shi-Qi , H. Bin-Jie , W. Xian-Yi , ‘’Hierarchical
Cooperative Spectrum Sensing Based on Double
Thresholds Energy Detection’’ , IEEE Communications
Letters,16, 7, (2012), pp.1096-1099.
[17] A T H m “T
largest eigenvalue in the gaussian ensembles," in
Calogero-Moser-Sutherland Models, J. van Diejen and
L. Vinet, eds., pp. 461-472. New York: Springer, 2000.
BIOGRAPHIES
Cebrail Ciflikli was born in K. Maras,
Turkey, in 1961. He received the Ph.D.
degree in electronics engineering from
Erciyes University in 1990. In 2004, he
joined Erciyes University Kayseri
Vocational College as Professor where
he is now Principal. Dr. Ciftlikli's
current research interests include
spread-spectrum communications,
wireless ATM/LAN, signal processing,
DS-CDMA system engineering, RF
power amplifier linearization for
wireless communication systems.
Fatih Yavuz Ilgin was born in Erzincan,
Turkey, in 1982. He received the B.S.
degree in Electronic And Computer
Education from the Gazi University,
Ankara, in 2007, the M.S. degree in
electrical and electronics engineering
from the Black Sea Technical
University, Trabzon, Turkey, in 2013.
In 2007 he joined the Erzincan
University Vocational High School,
Electronic and Automation
Department, Erzincan University

On the Performance Analysis of Blind Spectrum Sensing Methods for Different Communication Channels

More Related Content

What's hot (19)

Similar to On the Performance Analysis of Blind Spectrum Sensing Methods for Different Communication Channels (20)

More from IRJET Journal (20)

Recently uploaded (20)

On the Performance Analysis of Blind Spectrum Sensing Methods for Different Communication Channels