Content uploaded by Noor Gul
Author content
All content in this area was uploaded by Noor Gul on Feb 15, 2022
Content may be subject to copyright.
Proc. of the 1
st
International Conference on Electrical, Communication and Computer Engineering (ICECCE)
24-25 July 2019, Swat, Pakistan
978-1-7281-3825-1/19/$31.00 ©2019 IEEE
H. Asfandyar
Department of Electronics
University of Peshawar
Peshawar, Pakistan
hashmiasfandyar@yahoo.com
N. Gul
Department of Electronics
University of Peshawar
Peshawar, Pakistan
noor.phdee51@iiu.edu.pk
I.Rasool
Department of Electronics
University of Peshawar
Peshawar, Pakistan
imtiazrasoolkhan@uop.edu.pk
A. Elahi
Department of Electrical Engineering
International Islamic University
Islamabad, Pakistan
Atif.phdee40@iiui.edu.pk
Abstract—Spectrum sensing is of great importance in Cognitive
Radio Network (CRN). It cannot be achieved by a single user
owing to multipath fading and shadowing effect. Therefore, more
than one user is required to accurately sense the spectrum
availability as in Cooperative Spectrum Sensing. All sensing
statistics are collected at Fusion Center (FC) from cooperative
users and FC combines them to reach to a common global
decision. As these users are far apart from each other, therefore
they experience different channel conditions. So, it is necessary to
deal with the incoming data received from these Secondary Users
(SU) differently. The proposed scheme of Flower Pollination
Algorithm (FPA) intelligently finds optimum weighting
coefficients against cooperative users’ information and utilizes
these weights in the global decision of the Soft Decision Fusion
(SDF). This scheme is able to find optimum weights that lead to
minimum false alarm, high detection and minimum error
probability. The system is simulated for different numbers of the
cooperative users and Signal-to-Noise Ratios (SNRs) that shows
better sensing performance of the proposed FPA based scheme
against the Differential Evolution (DE), Genetic Algorithm (GA),
Maximum Gain Combination (MGC), Particle Swarm
Optimization (PSO), and Count techniques.
Index Terms—Cognitive Radio, Flower Pollination Algorithm,
Cooperative Spectrum Sensing
I. INTRODUCTION
n wireless communication system, the Electromagnetic
radio spectrum is considered as a rare resource. The
development of wireless technology and devices has
increased the demand of spectrum bands for communication.
According to a survey conducted by Federal Communication
Commission (FCC) the main cause of spectrum engagement is
the underutilization of channel by licensed user [1] . In order
to efficiently use the spectrum, Cognitive Radios (CRs) with
adaptive intelligence is getting the attention of researchers to
overcome such communication constrains. These nodes are
known as Secondary Users (SUs) or Unlicensed Users in
CRN [2].
The idea of CR is to periodically sense the communication
spectrum, detect spectrum availability and opportunistically
utilize the available resources without any interference to the
Primary User (PU). Energy Detection Scheme (EDS) is one of
the most optimal spectrums sensing technique in order to
detect the spectrum holes of PU regardless of knowing their
location, structure and strength. Nonetheless, after the
existence of shadowing and hidden terminal problem, the PU
signal might not be detected by the SU within the bounded
sensing time [3]. This means that EDS is highly exposed to the
channel effects like multipath fading and fluctuation due to
noise power. In [4], [5]. Cooperative Spectrum Sensing (CSS)
is proposed to overcome such channel effects in which the
PUs activity is observed by multiple SUs and to acquire the
band immediately if PUs absence has been detected.
The decision of all SUs is collected at a central point in CSS
known as Fusion Centre (FC). The FC decides about the
absence or presence of PU by combining all the reports from
SUs. These schemes are classified as Soft Decision Fusion
(SDF), Softened Hard Decision Fusion (SHDF), and Soft
Decision Fusion (SDF) [6]-[7]. A single bit decision is made
by SUs which is then forwarded to FC for further necessary
action in HDF scheme. On contrary, in the SDF scheme, the
readings of SUs work as a raw material for FC to make final
decision about the activity of PU. The results shown by the
SDF scheme are far better than HDF scheme [8], [9].
In the proposed work, all cooperative users employ energy
detector that compares received signal energy of the channel
with an adaptive threshold determined by the Flower
Pollination Algorithm (FPA). As the cooperative users in the
proposed work are considered at different geographical
locations and experience independent Raleigh fading effects,
therefore, it is not suitable to treat their sensing performances
equally in the global decision made by the FC. Similarly, in
the proposed method the FPA instead of keeping fixed
threshold point for all sensing intervals determines an
optimized threshold value. The weighted coefficient vector
with optimum threshold point is selected by the proposed
method. This leads to a minimum false alarm, high detection
and low error probability at the FC.
Enhanced Cooperative Spectrum Sensing in
Cognitive Radio Network Using Flower
Pollination Algorithm
I
Authorized licensed use limited to: University of Peshawar. Downloaded on September 14,2020 at 06:40:05 UTC from IEEE Xplore. Restrictions apply.
The proposed schemes in [10] - [11] determined coefficient
vectors by employing genetic algorithm (GA) and Particle
Swarm Optimization (PSO). In this paper, the optimal
coefficient weighted results are achieved using FPA. The final
weighted results are further utilized by the SDF to reach to a
final global decision at the FC. Simulation results are collected
for different number of SUs, Signal-to-Noise ratios (SNRs).
The results demonstrate a more sophisticated detection
performance by the proposed FPA based CSS as compared to
the PSO, GA, DE, MGC, and count schemes.
The rest of the paper is organized as following: Section II
elaborates the system mode. Section III explains the proposed
method in determining optimal weighted results using FPA.
Simulation results are demonstrated in Section IV. Finally, the
paper is concluded in Section V.
II. SYSTEM MODEL
The block diagram of the proposed system is illustrated in
figure. In this diagram, FC receives statistical observations of
the
M
SUs about the channel. The sensing users in the
diagram operate similar to a forwarding relay that simply
receive and forward the received PU signal to the FC. The
final verdict regarding the presence of the licensed user is
made at the FC using linearly weighted SDF based CSS that
use received signal information of the SUs.
The binary hypothesis about the presence and absence of the
PU activity observed by each user is as:
0
1
:[] [] , 1, 2,..., , 1, 2,...,
:[] [] []
i
i
ii i
HYn Wn iMnK
HYn gSnWn
=
∈∈
=+
(1)
Where
0
H
hypothesis shows no PU activity and
1
H
hypothesis
tells us about the occupancy of spectrum by PU. In the given
hypothesis,
[]
i
Yn
is the energy of the received signal of the
th
i
user at the
th
n
time slot. The total number of samples is
2
s
K
BT=
that are considered large enough to make the energy
distribution Gaussian. Here, B, is the signal bandwidth and,
,
s
T
is the sensing time. In (1),
,
i
g
is the gain of the channel
between the
th
i
user and PU and
[]Sn
is the
th
n
sensing samples
that are contemplated as an independent and identically
distributed (i. i. d) Gaussian random process whose mean is
zero and variance is given as,
2
,
S
σ
i.e.
()
2
[]~ 0,
S
Sn N
σ
.
[]
i
Wn
in the (1) is the Additive White Gaussian Noise (AWGN) of
channel between i
th
user and PU. Its mean is also zero and
variance is given as,
2
,
i
W
σ
i.e.
2
[]~ (0, )
i
iW
Wn N
σ
.
The final test statistic observed at the FC based on the
received signal of all cooperative users is made as
()
1
M
ii
i
Z
wZ
=
=
where
K2
n=1
U[n]
im
Z=
is the total energy
samples collected from the
th
i
user at the FC and
,
[] [] []
iRiiii
Un P hYn Nn=+
is the signal received at the FC
respectively.
Figure 1: The proposed CSS Model
Here,
,,
R
i
P
is the
th
i
user transmitting power to the FC and
i
h
is
the channel gain between the FC and i
th
sensing user. It is
further assumed that N
i
[n] is the AWGN between SU and FC.
Its mean is also zero and variance is represented by,
2
,
i
δ
i.e.
()
2
[]~ 0,
ii
Nn N
δ
. Similarly,
i
w
is the weight assigned to the
th
i
sensing user. As
i
Z
is normally distributed, therefore, the
resultant test statistic,
,
Z
is also follows normal distribution
[12] and [13].
()
2
00,
1
M
ii
i
EZH wK
σ
=
=
(2)
()
2
11,
1
M
ii
i
EZH wK
σ
=
=
(3)
()
0
2222
00,
1
var 2 ( )
MT
iii H
i
Z
HwK ww
σδ
=
=+=Φ
(4)
()
1
2222
11,0,
1
var 2 ( )
MT
iii H
i
Z
HwK ww
σσ
=
=+=Φ
(5)
Here,
22
0, 1,ii
and
σσ
are the variances of
[]
i
Un
under the
0
H
and
1
H
hypothesis made by the
th
i
user that are equivalent to
2
222
0, ,
i
iRiiWi
Ph
σσδ
=+
and
22
222
1, , 0,iRiiis i
Pg h
σσσ
=+
respectively.
In (2) to (5),
[]
12
,
T
M
ww w
w=
are the weighting
Authorized licensed use limited to: University of Peshawar. Downloaded on September 14,2020 at 06:40:05 UTC from IEEE Xplore. Restrictions apply.
coefficient vectors. These weights are then optimized to
determine the appropriate threshold value
β
.
The
0
H
and
1
H
hypothesis covariance matrices are
()
0
4
0,
2
Hi
diag K
σ
Φ=
and
()
1
222 22
,0,
2( | || | )
HRiiiSi
diag K P g h
σσ
Φ= +
respectively. In these matrices,
()
.diag
is a square diagonal
matrix whose rest of the entries are zero and only diagonal
elements are the elements of a given weighting vector. The
final results of the false alarm and detection probabilities at
the FC can be represented as:
()
0
00
0
0
()
var( )
T
fT
H
EZH w
PPZ H Q Q
ZH ww
ββμ
β
−−
=> = =
Φ
(6)
()
1
11
1
1
()
var( )
T
dT
H
EZH w
PPZ H Q Q
ZH ww
ββμ
β
−−
=> = =
Φ
(7)
Where,
11
01
01
TTTT
HH
TT
HH
wwwwww
wwww
μμ
β
Φ+Φ
=
Φ+Φ
Assuming that the
f
m
P
P=
, where
m
P
is the misdetection
probability and
1
f
d
P
P=−
, therefore, the total error
probability
e
P
is determined as:
01
01
TT
efm TT
HH
ww
PPP Q Q
ww ww
βμ
μβ
−−
=+= +
ΦΦ
(8)
In (8) the error probability is the fitness function and is highly
dependent on the selection of the
w
. Therefore,
β
is
optimized for the selection of the weighting coefficients and
substituting back into (8) leads to a high detection, minimum
false alarm, and low error probability. However, in the
proposed work selection of the
w
is perform in order to reduce
selection of the search space with
01
i
w<<
and
2
1
1
M
i
i
w
=
=
.
III. PROPOSED FLOWER POLLINATION ALGORITHM
BASED WEIGHTING METHOD
The FPA was developed by Xing-she Yang in 2012 inspired
by the pollination process of flowering plants. FPA has been
extended to multi-objective optimization with promising
results [14], [15].
In the proposed method, FPA finds the optimal set of weighted
coefficient vector against the sensing reports received from all
cooperative users. In the random normalized set of coefficient
vector population the vector with low error probability results
are elected as the optimal set of vector and is further utilized
in the global decision of the SDF scheme.
The steps involved in optimization process are given below:
Step 1: Initial Population
The algorithm initializes the initial population by randomly
generating N flower or pollen gametes i.e.
12
[ ] , 1,......,
T
M
ww w s N=∈
w. These values are
normalized between the range of 0 and 1.
Step 2: Fitness of the pollen gametes
It determines the suitability of each coefficient vector by
measuring their fitness scores
12
( ), ( ),...., ( )
ee eN
P
wPw Pw
. The
population is arranged in the increasing order of their fitness
measure.
Step 3: Global and Local Pollination
In this step, either global or local pollination is performed with
the help of current best solution and global best pollens. The
interaction or switching between global and local pollination
is controlled by probability switch p ϵ [0, 1] slightly biased
towards local pollination .This process results in new
population.
Step 4: New Population
The fitness of new population is determined in the same way
as described in step 2. The results are then sorted in ascending
order of their fitness and step 3 is repeated again.
Step 5: Stopping Criteria
FPA repeats step 2 time and again until the minimum P
e
is not
achieved or given number of iteration are not completed.
IV.
SIMULATIONS
AND
RESULTS
In the simulation, different cooperative users are initialized in
CRN with SNR varying from -25 dB to +10dB. The sensing
interval is selected 1 ms having 200 samples. Cooperative
users expressing different SNRs, sense the PU channel
independently. The size of the population for FPA is selected
consisting
M
number of pollens in each flower with total N
number of flowers. Total number of iteration is kept at 10. For
more promising results, probability switch is set at p = 0.8.
The proposed FPA scheme returns the optimum weighting
coefficient vector that is further used in defining a perfect
threshold value beta, β, for minimization of error probability.
Figure 2 and 3 depicts graph between error probability verses
increasing SNR having seven and twenty-one number of users
respectively, for proposed FPA, GA, PSO, DE, MGC and
count. It is clear from the figure that with increasing number
of users, error probability of all schemes decreases. The
graphical result illustrates that MGC-SDF has a worst
performance in the Raleigh fading environment while
detecting PU activity, followed by count, PSO, GA and DE.
The proposed FPA-SDF scheme in Figure 2 is able to detect
the channel with less error at all SNRs values compared with
other traditional schemes.
In Figure 3, the number of users has been increased from
seven to twenty-one, which shows that by increasing number
of users, results are getting more promising. In figure 4 and 5,
the graphs of the proposed FPA, GA, PSO, MGC DE and
count are plotted between error probability and increasing
number of users with fixed SNR. Results show that with
increasing number of total cooperative users, error probability
of these combination schemes decreases. The graphs further
illustrates that just like the case in Figure 2, the MGC scheme
Authorized licensed use limited to: University of Peshawar. Downloaded on September 14,2020 at 06:40:05 UTC from IEEE Xplore. Restrictions apply.
shows worst results followed by count, PSO, GA and DE. In
all these combination schemes our proposed method of the
FPA-SDF is able to detect the PU channel more accurately
that leads to lower detection error.
Figure 2: Probability of Error vs. Signal to Noise Ratios with Seven
Users
Figure 3: Probability of Error vs. Signal to Noise Ratios with
Twenty-One Users
Figure 4: Probability of Error vs. Total Cooperative Users with
Average SNR
Figure 5: Probability of Error vs Total Cooperative Users with
Average SNR
V.
CONCLUSION
The fading and shadowing effects due to Raleigh fading
channel reduces the sensing performance of an individual user.
Proposed FPA based CSS in the paper is able to determine
optimal coefficient vectors against the reporting users before
SDF scheme is allowed to take a global decision at the FC.
The optimal coefficient vector is able to produce high
detection, minimum false alarm and low error probability for
the proposed FPA-SDF scheme compared to the PSO-SDF,
DE, GA, count and MGC-SDF schemes at varying SNRs and
cooperative user’s participation.
R
EFERENCES
[1] and G. D. R. Engelman, K. Abrokwah, “Federal
Communications Commission Spectrum Policy Task
Force Report of the Spectrum Efficiency Working
Group,” Fed. Commun. Comm. Spectr. Policy Task
Force, vol. 1, no. 1, p. 37, 2002.
[2] N. Gul, I. M. Qureshi, S. Akbar, M. Kamran, and I.
Rasool, “One-to-Many Relationship Based Kullback
Leibler Divergence against Malicious Users in
Cooperative Spectrum Sensing,” Wirel. Commun.
Mob. Comput., vol. 2018, pp. 1–14, 2018.
[3] N. Gul, I. Mansoor, Q. Aqdas, N. Atif, and E. Imtiaz,
“Secured Soft Combination Schemes Against
Malicious Users in Cooperative Spectrum Sensing,”
Wirel. Pers. Commun., no. 0123456789, 2019.
[4] W. Zhang, R. K. Mallik, and K. Ben Letaief,
“Cooperative spectrum sensing optimization in
cognitive radio networks,” IEEE Int. Conf. Commun.,
pp. 3411–3415, 2008.
[5] C. Qi, J. Wang, and S. Li, “Weighted-clustering
cooperative spectrum sensing In cognitive radio
context,” Proc. - 2009 WRI Int. Conf. Commun. Mob.
Comput. C. 2009, vol. 1, no. 2007, pp. 102–106, 2009.
[6] A. Rauniyar and S. Y. Shin, “Improved Detection
Performance of Energy Detector by Optimization of
Threshold Using BPSO Algorithm for Cognitive
-25 -20 -15 -10 -5 0 5 10
SNR
0
0.1
0.2
0.3
0.4
0.5
0.6
FPA
Count
MGC
PSO
GA
DE
10 15 20 25 30
Number of Users
0.25
0.3
0.35
0.4
0.45
0.5
0.55
0.6
DE
PSO
FPA
GA
Count
MGC
Error Probability
Authorized licensed use limited to: University of Peshawar. Downloaded on September 14,2020 at 06:40:05 UTC from IEEE Xplore. Restrictions apply.
Radio Networks,” pp. 179–183, 2015.
[7] M. Emami, H. Zarrabi, M. R. Jabbarpour, M. Sadat
Taheri, and J. J. Jung, “A soft cooperative spectrum
sensing in the presence of most destructive smart
PUEA using energy detector,” Concurr. Comput., vol.
30, no. 15, pp. 1–10, 2018.
[8] Z. Quan, W. Ma, S. Cui, and A. H. Sayed, “jH jH,”
Spectrum, vol. 58, no. 4, pp. 2431–2436, 2010.
[9] M. Akbari, M. R. Manesh, S. A. R. T. Zavareh, and P.
Shahabi, “Maximizing the Probability of Detection of
Cooperative Spectrum Sensing in Cognitive Radio
Networks,” Prog. Electromagn. Res. Symp.
Proceedings, KL, MALAYSIA, pp. 123–126, 2012.
[10] M. Akbari, M. R. Manesh, A. A. El-Saleh, and M.
Ismail, “Improved soft fusion-based cooperative
spectrum sensing using particle swarm optimization,”
IEICE Electron. Express, vol. 9, no. 6, pp. 436–442,
2012.
[11] N. Gul, I. M. Qureshi, A. Elahi, and I. Rasool,
“Defense against Malicious Users in Cooperative
Spectrum Sensing Using Genetic Algorithm,” Int. J.
Antennas Propag., vol. 2018, pp. 1–11, 2018.
[12] M. Akbari and M. Ghanbarisabagh, “A Novel
Evolutionary-Based Cooperative Spectrum Sensing
Mechanism for Cognitive Radio Networks,” Wirel.
Pers. Commun., vol. 79, no. 2, pp. 1017–1030, 2014.
[13] N. Gul, I. M. Qureshi, A. Omar, A. Elahi, and S.
Khan, “History based forward and feedback
mechanism in cooperative spectrum sensing including
malicious users in cognitive radio network,” PLoS
One, vol. 12, no. 8, pp. 1–21, 2017.
[14] A. A. . El-Saleh, M. . Ismail, M. A. M. . Ali, and M.
K. . Hossain, “Biologically-inspired soft fusion
scheme for cooperative spectrum sensing in cognitive
radio networks,” 7th Int. Conf. Inf. Technol. Appl.
ICITA 2011, no. Icita, pp. 239–244, 2011.
[15] X. Yang, Nature-Inspired Optimization Algorithms. .
Authorized licensed use limited to: University of Peshawar. Downloaded on September 14,2020 at 06:40:05 UTC from IEEE Xplore. Restrictions apply.