Wireless Pers Commun (2015) 82:21712184

DOI 10.1007/s11277-015-2340-3

Optimal Cooperative Spectrum Sensing in

Interference-Aware Cognitive Radio Networks
Fei Song Changju Kan Qihui Wu Guoru Ding

Published online: 3 February 2015

Springer Science+Business Media New York 2015

Abstract In order to protect the primary user from being interfered, most of the related
works focus only on the restriction of the missed detection probability, which may cause
over-protection of the primary user. Thus the interference probability is defined and the
interference-aware sensing model is introduced in this paper. The interference-aware sensing
model takes the spatial conditions into consideration, and can further improve the network
performance with a good spectrum reuse opportunity. Meanwhile, although the detection
performance can be further improved by the cooperative spectrum sensing, it also introduces
additional reporting time corresponding to the number of cooperative users (CUs), which
may decrease the throughput of the CRN. Motivated by the above, in this paper, we study
the throughput tradeoff for interference-aware cognitive radio networks. For cooperative
spectrum sensing, the Logic-OR fusion rule is used. By jointly optimizing the sensing time
and the number of the CUs, the maximum throughput can be achieved. Theoretical analysis
is given to prove the feasibility of the optimization and numerical simulations also show that
the maximum throughput can be achieved when the sensing time and the number of the CUs
are both optimized.
Keywords Cognitive radio networks Cooperative spectrum sensing Interference-aware
Sensing overhead Throughput optimization

1 Introduction
A severe problem caused by the popularity of wireless communication is the lack of spectrum resources, which is introduced by current legacy command-and-control regulations [1].
However, it is known to all that the assigned spectrums are underutilized in both spatial
and temporal dimensions. Therefore, the spectrum scarcity results from the fixed spectrum
assignment policy rather than the physical scarcity of spectrum resources. In order to solve

F. Song C. Kan (B) Q. Wu G. Ding

College of Communications Engineering, PLA University of Science and Technology, Nanjing, China
e-mail: kanchangju@163.com

123

2172

F. Song et al.

the problem of spectrum scarcity, cognitive radio networks (CRNs) has been proposed to
overcome this problem which allows dynamic spectrum reuse [25]. Since detecting the
spectrum hole is an essential method to reuse the registered spectrum, the spectrum sensing
function becomes one of the key technologies of the cognitive radio [6,7].
Spectrum sensing, as a fundamental problem in CR, requests the secondary user (SU) to
efficiently and effectively detect the presence of the primary user (PU) [8,9]. However, due to
many environmental factors such as low signal-to-noise radio (SNR), multi-path fading and
shadowing, the sensing performance may be inherently limited, which makes the spectrum
sensing problem more involved. In order to further improve the sensing performance, cooperative spectrum sensing (CSS) has been studied extensively [1015]. In CSS, cooperative
users (CUs) individually sense the channels, send information to the secondary user and then
through proper fusion of the collected information, the SU will make the final decision. There
are various cooperative sensing schemes to fuse the sensing information of the secondary
users. The schemes can be classified into hard decision based fusion, soft decision based
fusion [9] and data based fusion schemes [10]. In this paper, we consider the Logic-OR rule,
as it achieves the highest detection probability and is easy to implement.
In order to analyze the performance of the spectrum sensing, two basic parameters, detection probability and false alarm probability, are widely used and accepted by the world
[1619]. The higher the detection probability gets, the better the PU can be protected from
the interference of SU. Meanwhile, the lower the false alarm probability is, the more chances
the SU can have to reuse the registered channel. Thus, a fundamental tradeoff is appeared
between the two probabilities. In order to improve the spectrum sensing performance, several
system models aiming at optimizing the tradeoff are established and are widely accepted for
spectrum sensing. Specifically, reference [19] designed a frame structure and held the classical idea that a longer sensing time will get a higher detection probability as will as a lower
false alarm probability. But within a fixed frame size, the longer sensing time will shorten
the data transmission time of the secondary users. Thus, an optimal tradeoff of sensing and
throughput is investigated in [19]. However, reference [19] ignored the influences made by
the spatial environment. In fact, the distance between PU and SU may have an impact on the
sensing performance. Based on this idea, reference [20] introduced a new concept named
interference-aware spectrum sensing that taken the distance between PU and SU into the consideration. It argued that even though the SU makes a missed detection, there still exists the
case that SU does not interfere with PU due to the actual spatial distances between PU and SU.
Finally, the sensing performance is well analyzed in this interference-aware sensing model.
Meanwhile, as analyzed in [19], sensing time is an important component in the design
of cognitive radio networks and is optimized in the sensing-throughput tradeoff. In our
previous work, we have investigated the sensing-throughput tradeoff in cooperative spectrum
sensing. Also, the influence produced by imperfect reporting channel is analyzed. However,
in cooperative spectrum sensing, except for the sensing time, each CU has to spend some
reporting time for sending its sensing information to the SU. The reporting time, which
is called the cooperative overhead and related with the number of cooperative users, may
decrease the transmission time and finally reduce the throughput of the CRN. Thus, the
reporting time in CSS is also an important factor and should be considered in the optimization
of the CSS for a CRN.
Motivated by the above considerations, in this paper we firstly introduce the interferenceaware spectrum sensing model. Then by using the Logic-OR fusion rule as the basis, and
both the sensing time and the cooperative overhead are considered, the issue of sensingthroughput tradeoff in interference-aware cognitive radio networks is investigated. Finally,
the optimization problem of the tradeoff is formulated and the achievable throughput is

123

Optimal Cooperative Spectrum Sensing

2173
Reporting
Slot

PU is active

PU is inactive

Sensing 1
Slot

Transmission
Slot

Sensing
Slot

SensingPeriod
CU1
CU2

PU

CUi

d
CU3

SU

Fig. 1 System model of cognitive radio network (R: radius of PU; d: the distance between PU and SU)

maximized by jointly optimizing the sensing time and the fusion parameter N along the
distance between PU and SU. The analytical and numerical results obtained in this paper
clearly show that the maximum throughput can be achieved when the sensing time and the
number of CUs are both optimized.
The paper is organized as follows. Section 2 presents the system model for spectrum sensing in interference-aware cognitive radio networks. Section 3 investigates the optimization
problem and then figures out the existence of optimal sensing time and N value. Section 4
presents some theoretical analysis of the optimization problem and gives solution for the
interference-aware spectrum sensing. Performance evaluation and comparisons are given in
Sect. 5. And finally, conclusions are drawn in Sect. 6.

2 System Model
In this section, we first present the system model of the interference-aware cognitive radio
networks. In this part, a brief introduction of the special cognitive radio network is given. Then
the local sensing model which considering the energy detection scheme and the cooperative
sensing model which considering the Logic-OR fusion rule are presented in the following.
At last, the probability of interference is defined.
2.1 Network Model
As can be seen from Fig. 1, we consider a cognitive radio network where a SU is looking for
a chance to access the registered spectrum band. Around the SU, several cooperative users
are performing the sensing process to help the SU for the final decision. Denote R as the
radius of PU and the d as the distance between PU and SU. A synchronous system is assumed
and a frame structure of periodic spectrum sensing is presented. In each sensing period T,
we further divide the sensing period into three slots, the sensing slot, the reporting slots and
the transmission slot. During the sensing slot, each cooperative user performs its spectrum
sensing individually, then reports the sensing result to the SU orderly during the reporting
slot, and finally the SU determines the state of PU based on the spectrum sensing information
of each CU.
Because of the complicated environment factors, the CUs make mistakes during spectrum
sensing inevitably. If the PU is active while SU makes a missed detection, the SU will have the

123

2174

F. Song et al.

opportunity to use the frequency band. However, this behavior may bring interference to PU
related to the distance d. Thus, the probability of interference is taken into consideration with
the missed detection made by SU to further analyze the effects SU made on PU. The specific
analysis of the probability of interference will be showed in the following part. The main
work of this paper is to maximize the performance of spectrum sensing in interference-aware
cognitive radio networks.
2.2 Sensing Model
For each CU, the energy detection scheme is proposed. In energy detection, none of the prior
knowledge about the PUs signal is necessary, and the only need is to compare the energy of
the received signal with a pre-defined threshold in order to obtain a decision on the presence
of PU. Suppose yi (n) represents the received signal of SUi during the sensing time, and
then the PUs detection problem can be figured out as a binary hypothesis test between the
following two hypotheses.

H1 :yi (n) = xi (n) + w(n)
n = 1, 2, . . . K i
(1)
H0 :yi (n) = w(n)
where H0 and H1 denote that the PU is absent and present respectively. xi (n) is the PU signal
received at the CUi. w(n) is the background noise. Here we assume that the background
noise is AWGN and the PU signal is a Gaussian signal.
As the CUs make individual sensing decision, let i be the detection threshold, si the
sensing time, f s the sampling frequency and Ki the number of samples (Ki = si f s ) of the
CU i. If the number of samples K is adequately large (e.g. K  10), we can use the central
limit theorem to simplify the mathematic model, thus the probabilities of false alarm and
missed detection for each CU can be defined and calculated as follows [20]:



i
si f s
(i)
P f (si , i ) = Q
1
(2)
2
2



i
si f s
(i)

1
PMD (si , i ) = 1 Q
(3)
2
P + 2
where P is the average received power of the primary signal by SU and 2 is the AWGN
variance.
2.3 Cooperative Spectrum Sensing
CSS can address problems posed by low SNR, shadowing, and fading. In this paper, we
consider the Logic-OR decision fusion rule in CSS. Under the Logic-OR fusion rule, each
CU makes a binary decision based on its local observation and then forwards the decision
to the SU through the reporting channel. At the SU, all the decisions are fused together
according to logic rule.
We assume that, compared with the distance from any CU to the PU, the distance between
any two CUs is small, so that the received signal at each CU experiences almost identical
path loss. Hence, the CUs are assumed to performing same performance. Thus this results
in that the false alarm probability and the detection probability of each CU is independent
of each other. Let P f denote the false alarm probability, Pd denote the detection probability.
Therefore, based on the Logic-OR fusion rule, the final probability of false alarm and the
final probability of detection after fusion are given as follows:

123

Optimal Cooperative Spectrum Sensing

Q d (N , s , ) = 1

2175



1 Pdi s,i , = 1 (1 Pd (s , )) N

(4)

i=1

Q f (N , s , ) = 1




N
1 P if s,i , = 1 1 P f (s , )

(5)

i=1

2.4 Probability of Interference

As can be seen from Fig. 1, we consider the worst case in this paper that the primary receiver
(PR) just lies in the intersection of the primary transmitter (PT) coverage boundary and the
line connecting the PT and the SU. Whenever SU makes a missed detection, if SU lies within
the radius of PT, then the SU does not have the access to the registered frequency band or
it will cause interference to the PU without any doubt. Else if ST lies outside the radius of
PT, we can figure out the PRs received SNR by denoting PP as the PU power received
by PR and PC as the power of secondary signals received by PR. Only when the case occurs
that the received SNR of PR is smaller than the desired SNR t , rather than any missed
detection, can we draw a conclusion that the PU is interfered by the SU. Thus the probability
of interference can be defined and calculated as [20]

PI = Prob ( < t ) = Q


PP
PC

2
M

PC 2
PP2
t2

+ PC2 +

(6)

where M is the number of symbols in one packet during the reception.

3 Problem Formulation
In this section, in order to maximize the average throughput of cognitive radio networks, we
jointly consider the problems of spectrum sensing parameter setting and CU assignment in
cooperative spectrum sensing. To predefined, let s be the sensing time of each CU during
the sensing process, r be the reporting time of each CU to the SU. Then the optimization
problem is formulated under decision fusion rule as the follows.
There are two scenarios for which the secondary network can operate at the registered
channel:
1. When the PU is absent and no false alarm is generated by SU. In this scenario, except for
the wasted sensing time and the reporting time, the achievable throughput of secondary
networks is figured as
R0 (N , s , ) =



T s N r
P (H0 ) 1 Q f (N , s , ) C0
T

(7)

2. When the PU is present but is estimated to be absent by SU (missed detection). In this

scenario, an interference that made by SU to PU is engendered inevitably. However,
taking the spatial conditions into consideration, the influence of SU to PU also changes
along with the variation of the distance between the PU and the SU. Thus, avoiding the

123

2176

F. Song et al.

overly protection, the achievable throughput of this scenario is expressed as

R1 (N , s , ) =

T s N r
P (H1 ) (1 Q d (N , s , )) (1 PI ) C1
T

(8)

Let PS be the power of secondary signals received by the SU and N0 be the noise power.
Then can be respectively expressed as C0 and C1 , the specific formula is written as follows:


PS
C0 = log2 1 +
(9)
N0


PS
(10)
C1 = log2 1 +
P + N0
Thus the average throughput of the cognitive radio network is given as
max R (N , s , ) = R0 (N , s , ) + R1 (N , s , )

(11)

s.t.Q MD PI

(12)

N ,s ,

From (11) and (12) we can see that the achievable throughput is the function of the sensing
time s , the number of CUs N and the detection threshold . By the constraint condition that
the missed detection probability Q MD is satisfied, we are able to determine a threshold with
a certain N value and the sensing time s .





2
2
1
= P +
Q (1 PMD )
1+
s f s








2
N
2
1
= P +
1 Q MD
Q
1+
(13)
s f s
Thus, by combining the formula (2)(5), it is clear that the optimal goal is the function of
the parameter N value and sensing time s . So, the main work in this paper is to maximize
the achievable throughput by jointly optimizing the sensing time and the N value of the
Logic-OR fusion rule in cooperative spectrum sensing.
Then the average achievable throughput of the cognitive radio network is reduced to
max R (N , s ) = R0 (N , s ) + R1 (N , s )
N ,s



T s N r 
C0 P (H0 ) 1 Q f + C1 P (H1 ) (1 Q d ) (1 PI )
T
(14)
s.t.Q MD PI

(15)

4 Performance Analysis
In this section, we present some theoretical analysis of the optimization problem, in which a
lemma and two propositions are given. Based on the analysis, the joint optimization algorithm
is proposed.
Lemma 1 d, the maximum throughput is achieved with equality constraint in (15).

123

Optimal Cooperative Spectrum Sensing

2177

Proof For a given distance d, if a given sensing time s and a N value is given,
then let be the particular threshold that is certain to satisfy the constraint condition Q MD (N , s , )PI = , for any other threshold  that satisfies  , we have
Q d (N , s ,  )PI Q d (N , s , )PI ,thus Q MD (N , s ,  )PI which meets the condition in (15). However, from (7) and (8) we can deduce that R0 (N , s ,  ) R0 (N , s , )
and that R1 (N , s ,  ) R1 (N , s , ), so R(N , s ,  ) R(N , s , ). This proves that the
maximum throughput is achieved only with the equality constraint in (15).
Proposition 1 Under the interference-aware condition that described in (14) and (15), for
a given N value, at any distances between PU and SU, there exists an optimal sensing time
in the range of [0, T N r ], which yields the maximum achievable throughput for the CRN.
Proof From lemma 1, we get the optimal condition that Q MD PI = . Meanwhile, because
of the implicit constraint that 0 Q MD 1, we divide Q MD into the following two cases:
Case 1: If /PI < 1, then the optimal value of Q MD can be written as Q MD = /PI .
According to (14), we have



T s N r 
C0 P (H0 ) 1 Q f + C1 P (H1 ) (1 Q d ) (1 PI )
T





T s N r
(16)
C0 P (H0 ) 1 Q f + C1 P (H1 ) (1 PI )
=
T
PI

R (s ) =

For a fixed distance, we can get the differential equation from (16) that



R
1 
C0 P (H0 ) 1 Q f + C1 P (H1 ) (1 PI ) (1 Q d )
=
s
T
Q f
T s N r

C0 P (H0 )
T
s
N 1 P f

Q f
= N 1 Pf
s
s





2
Pf
P
1

f
1
P

s
s

= exp
+ 1 + 2 Q 1 1
s
2
2 2

PI
2


fs 1 P

2 2 s 2

(17)
(18)

(19)

From the above we can get that

lim

s 0 s

lim

= + > 0

s T N r

(20)





R
1
=
C0 P (H0 ) 1 Q f + C1 P (H1 ) (1 PI )
<0
s
T
PI

(21)

Up to now, a conclusion can be reached that there exists an optimal sensing time to obtain
the maximum achievable throughput within interval (0, T N r ). Then exhaustive search is
needed to help finding the optimal sensing time, by which the maximum achievable throughput can also be calculated.
Case 2: If /PI 1, then the optimal value of Q MD can be written as Q MD = 1. According
to (14), we have

123

2178

F. Song et al.

R (s ) =



T s N r 
C0 P (H0 ) 1 Q f + C1 P (H1 ) (1 PI )
T

(22)

We can see that it is a special case of (16). Thus we have




1 
R
C0 P (H0 ) 1 Q f + C1 P (H1 ) (1 PI )
=
s
T
Q f
T s N r

C0 P (H0 )
(23)
T

N 1 P f

Q f
= N 1 Pf
(24)
s
s






 2
Pf
1
s f s P
fs 1 P
1
1
P

= exp
+
1+ 2
(25)

s
2
2 2
2

2 2 s 2
2
Obviously, we can get that

lim

s 0

lim

s T N r

R
= + > 0
s



1 
R
C0 P (H0 ) 1 Q f + C1 P (H1 ) (1 PI ) < 0
=
s
T

(26)
(27)

Thus, there is a maximum point of R( ) within interval (0, T N r ). Also, we can get the
best sensing time by the exhaustive search, as well as the maximum achievable throughput.
Proposition 2 Under the interference-aware condition that described in (14) and (15), for
a given sensing time s , at any distances between PU and SU, there exists an optimal N value
which yields the maximum achievable throughput for the CRN.
Proof When the N value is small, i.e., only few number of CUs takes part in the sensing
process, due to the complicated channel propagation condition, the final sensing performance
may not be satisfied, thus the advantage of cooperative sensing is not represented. However,
if the N value is large to some degree, although the sensing performance can be improved
because of the large CUs cooperation, the sensing overhead is also enlarged. As the whole
reporting time is longer, the transmission time get reduced correspondingly, thus the final
CRN performance may suffer a decrease. Considering the above, the selection of the N value
is important in the CSS. As a result, for a fixed distance, an optimal N value is existed for
the maximum throughput of CRN.
There is no closed-form solution for the optimal N value in this optimization. However,
since the number of the CUs N is an integer, it is not computationally expensive to search
the optimal N value that maximizes the achievable throughput.
Since the optimal sensing time s and the optimal number of CUs can be achieved and is
proved to be feasible by the above theoretical analysis, we summarize the proposed iterative
algorithm in Algorithm 1 as follows.

123

Optimal Cooperative Spectrum Sensing

2179

5 Performance Evaluation
In this section, numerical simulation results are presented to evaluate the throughput tradeoff
in interference-aware cognitive radio networks. Logic-OR fusion rule is used for final decision
and each cooperative user is assumed to use the energy detector for local sensing. Simulations
are carried out to find the optimal N value and sensing time s at each distance between the
PU and SU which achieve the maximum throughput while providing sufficient protection to
the primary user simultaneously.
Let the radius of primary cell be 500 m and the probability of activity PU be P(H1 ) = 0.5.
The transmit power of PU and SU are respectively 30 and 20 dBm, the noise variance is set
to 100 dBm. The bandwidth of the PU is set to 30 kHz, the frame duration is T = 20 ms,

123

2180

F. Song et al.

Optimal numbers of CU

15
optimal number of CUs

10

0
500

1000

1500

2000

Distance (m)

Fig. 2 Optimal number of CUs that maximize the achievable throughput

the reporting time of single CU is r = 0.2 ms. The constraint for protection is = 0.05 and
the desired SNR of the PU is t = 8 dB.
Figure 2 describes the optimal N Value in the Logic-OR fusion rule cooperative spectrum
sensing at different distances. As explained in the above, few CUs participating in the CSS
cannot get higher sensing performance, while too much CUs will raises the sensing overhead,
which can also decrease the CRN performance. Based on the searching algorithm, the optimal
problem can be solved and the optimal N values are achieved along the distance between
the PU and the SU. From the figure we can see that there is no single N value that meets
the optimal problem for all the distances, thus, an optimal N value is needed to achieve the
maximum throughput of the cognitive radio network.
Figure 3 indicates the variation of the optimal sensing time under different N values.
For different N value, the optimal sensing time is different correspondingly. Based on the
interference-aware sensing frame model, the selection of sensing time should be taken into
consideration, too. A short sensing time may cause lower detection performance, hence
decreasing the final CRN throughput. If the sensing time is longer to some extent, although
the detection performance is higher, the transmission time is shorter, which can also decrease
the final achievable throughput. Thus, an optimal sensing time also exists for the maximum
achievable throughput. From the figure we can see that the optimal spectrum sensing time is
different from each other along the distance between the PU and the SU. In order to solve
the optimization problem, an optimal N value is needed to achieve the maximum throughput
of the cognitive radio network.
According to the pair of N value and the sensing time s , then the special achievable
throughput is calculated. Thus, if the optimal N value and optimal sensing time s is gotten,
the maximum throughput should be achieved. The numerical results will be provided to
certify the conclusion in the following part.
Figure 4 shows the maximum achievable throughput of optimal sensing time at each
distance. The cases when the sensing time is fixed at s = 0.4 ms and s = 10 ms are
compared in Fig. 4. As a common condition, the K value is optimized in the three cases
along the distance between the PU and the SU. From Fig. 4 we can see that when the SU
is quite near to PU (e.g. d 800 m), the sensing performance is quite good as Q f is much
too small, so we should decrease the sensing time to get longer transmission time for higher
CRN throughput rather than further lessen Q f ; however, when the SU goes far away from
the PU (e.g. d 1,580 m), SU is almost interference-free to PU, which means that a higher
Q f can be endured, so we should also try to reduce the sensing time in order to prolong the

123

Optimal Cooperative Spectrum Sensing

2181

Optimal sensing time (ms)

15
Optimal N value
N=2
N=20

10

0
500

1000

1500

2000

1500

2000

Distance (m)

Achievable throughput R( ) (bit/s/Hz)

Fig. 3 Optimal sensing time of different N values

10
9
8
7

Optimal sensing time

Sensing time

= 0.4ms

Sensing time

= 10ms

6
5
4
3
2
1
500

1000

Distance (m)

Fig. 4 The achievable throughput of different sensing time

transmission time; when the distance lies between the two cases, as Q f increases, longer
sensing time should be taken when consider the overall throughput which is influenced by
both the false alarm probability and the transmission time. So, detailed theoretical analysis
is needed so as to search for the optimal sensing time. Thus, we can draw a safe conclusion
that when compared to the fixed sensing time cases, the proposed algorithm can optimize
the sensing time to maximum achievable throughput at each distance, for which the CRN
performance can be further improved.
Figure 5 shows the maximum achievable throughput of optimal N value at each distance.
Also, the cases when N is fixed to 2 (represents the small number of CUs case) and N fixed
to 20 (represents the large number of CUs case) are compared in Fig. 5. As a common
condition, the sensing time s is optimized in the three cases along the distance between the
PU and the SU. From the figure, we can see clearly that for different N values, the achievable
throughput is varied from each other. When the SU is close to the PU (d 800 m), the
local spectrum sensing of single CU can reach a satisfied result, so we should decrease
the number of CUs in order to reduce the sensing overhead, which means to prolong the
transmission time and then improve the CRN throughput; when the distance is far from the
PU (d 1,580 m), SU is almost interference-free to PU, which means that the constraint
condition of the spectrum sensing is enlarged, a small number of CUs can reach the desired
sensing performance, so we should also try to reduce the sensing overhead in order to prolong

123

F. Song et al.
Achievable throughput R( ) (bit/s/Hz)

2182
10
9
8
7

Optimal N value
number of CUs N = 2
number of CUs N = 20

6
5
4
3
2
1

500

1000

1500

2000

Distance (m)

Fig. 5 The achievable throughput of different N values

the transmission time and further achieve the CRN throughput. However, in the distances
among the range 800 m d 1,580 m, the sensing performance need to be insured by more
CUs participating in the CSS, we should consider the sensing overhead the more CUs can be
brought simultaneously, and then an optimal N is calculated to satisfy the special problem.
As a result, the proposed algorithm which optimizing the N value can reach a maximum
achievable throughput when compared the fixed N value cases.
Thus, in order to analyze the performance of cognitive radio networks, both the sensing
time and the parameter N value of the Logic-OR fusion rule should be taken into consideration.
By jointly optimizing the sensing time and the N value, we will finally get the maximum
achievable throughput in interference-aware cognitive radio networks.

6 Conclusion
In this paper, we studied the issue of throughput tradeoff problem of cooperative spectrum
sensing in interference-aware cognitive radio networks. For the cooperative spectrum sensing,
the Logic-OR fusion rule is used. By jointly optimizing the sensing time and the N value, the
maximum throughput can be well achieved in interference-aware cognitive radio networks.
The theoretical analysis and computer simulations are also given to show the capability of
improvement in CRN throughput of our proposed algorithm. Future interesting directions
include that taking the energy efficient problem into consideration and optimizing the problem
in the interference-aware cognitive radio networks.
Acknowledgments This work was supported by the National Science Foundation of China under Grant
Nos. 61172062 and 61301160, and in part by Jiangsu Province Natural Science Foundation of China under
Grant No. BK2011116.

References
1. Haykin, S. (2005). Cognitive radio: Brain-empowered wireless communications. IEEE Journal on
Selected Areas in Communications, 23(2), 201220.
2. Akyildiz, I. F., Mehmet, W.-Y. L., Vuran, C., et al. (2006). NeXt generation/dynamic spectrum
access/cognitive radio wireless networks: A survey. ELSEVIER Computer Networks, 50, 21272159.

123

Optimal Cooperative Spectrum Sensing

2183

3. Xu, Y., Shen, A., Shen, L., Wu, Q., et al. (2013). Decision-theoretic opportunistic spectrum access:
Strategies, challenges and solutions. IEEE Communication Survey & Tutorial, 15, 16891713.
4. Xu, Y., Wang, J., Wu, Q., et al. (2012). Optimal energy-efficient channel exploration for opportunistic
spectrum usage. IEEE Wireless Communications Letters, 1(2), 7780.
5. Xu, Y., Wang, J., Wu, Q., et al. (2012). Opportunistic spectrum access in cognitive radio networks: Global
optimization using local interaction games. IEEE Journal of Selected Topics in Signal Processing, 6(2),
180194.
6. Tandra, R., Sahai, A., & Mishra, S. M. (2009). What is a spectrum hole and what does it take to recognize
one. Proceedings of IEEE, 97(5), 824848.
7. Wu, Q., Ding, G., Wang, J., et al. (2013). Spatial-temporal opportunity detection in spectrumheterogeneous cognitive radio networks: Two-dimensional sensing. IEEE Transactions on Wireless Communications, 12(2), 516526.
8. Yucek, T., & Arslan, H. (2009). A survey of spectrum sensing algorithms for cognitive radio applications.
IEEE Communications Surveys and Tutorials, 11(1), 116130.
9. Zeng, Y., Liang, Y.-C., Hoang, A. T., et al. (2010). A review on spectrum sensing for cognitive radio:
Challenges and solutions. EURASIP Journal on Advances in Signal Processing, 2010, 116.
10. Quan, Z., Cui, S., & Sayed, A. H. (2010). Optimal linear cooperation for spectrum sensing in cognitive
radio networks. IEEE Journal of Selected Topics in Signal Processing, 2(1), 2840.
11. Ma, J., Zhao, G., & Li, Y. (2008). Soft combination and detection for cooperative spectrum sensing in
cognitive radio networks. IEEE Transactions on Wireless Communications, 7(11), 45024507.
12. Sayed, A. H., Cui, S., & Quan, Z. (2008). Optimal linear cooperation for spectrum sensing in cognitive
radio networks. IEEE Journal of Selected Topics in Signal Processing, 2(1), 2840.
13. Maleki, S., Chepuri, S. P., & Leus, G. (2011). Energy and throughput efficient strategies for cooperative
spectrum sensing in cognitive radios. In IEEE international workshop on signal processing advances in
wireless communications (pp. 7175).
14. Huogen, Y., Wanbin, T., & Shaoqian, L. (2011). Optimization of cooperative spectrum sensing in multiplechannel cognitive radio networks. In IEEE GLOBECOM.
15. Fan, R. F., & Jiang, H. (2010). Optimal multi-channel cooperative sensing incognitive radio networks.
IEEE Transactions on Wireless Communications, 9(3), 11281138.
16. Zhang, T., Wu, Y., Lang, K., et al. (2010). Optimal scheduling of cooperative spectrum sensing in cognitive
radio networks. IEEE Systems Journal, 4(4), 535549.
17. Zhang, W., Mallik, R. K., & Letaief, K. B. (2009). Optimization of cooperative spectrum sensing with
energy detection in cognitive radio networks. IEEE Transactions on Wireless Communications, 8(12),
57615766.
18. Peh, E., Liang, Y.-C., Guan, Y. L., et al. (2009). Optimization of cooperative sensing in cognitive radio
networks: A sensing-throughput tradeoff view. IEEE Transactions on Vehicular Technology, 58(9), 5294
5299.
19. Liang, Y., Zeng, Y., Peh, E. C. Y., et al. (2008). Sensing-throughput tradeoff for cognitive radio networks.
IEEE Transactions on Wireless Communications, 7(4), 13261337.
20. Lin, Y., Liu, K., & Hsieh, H. (2013). On using interference-aware spectrum sensing for dynamic spectrum
access in cognitive radio networks. IEEE Transactions on Mobile Computing, 12(3), 461474.

Fei Song received her B.S. degree in communications engineering,

and her Ph.D. degree in communications and information system from
Institute of Communications Engineering, PLA University of Science
and Technology, Nanjing, China, in 2002 and 2007, respectively. She
is currently a lecturer of PLA University of Science and Technology.
Her current research interests are cognitive radio networks, MIMO and
statistical signal processing.

123

2184

F. Song et al.

Changju Kan received his B.S. degree in communications engineering, from Institute of Communications Engineering, PLA University of
Science and Technology, Nanjing, China, in 2011. He is currently pursuing his M.S. degree in communications and information system in
Institute of Communications Engineering, PLA University of Science
and Technology. His research interests are cognitive radio networks and
green communication.

Qihui Wu received his B.S. degree in communications engineering, M.S. degree and Ph.D. degree in communications and information systems from Institute of Communications Engineering, Nanjing,
China, in 1994, 1997 and 2000, respectively. From 2003 to 2005, he
was a Postdoctoral Research Associate at Southeast University, Nanjing, China. From 2005 to 2007, he was an Associate Professor with
the College of Communications Engineering, PLA University of Science and Technology, Nanjing, China, where he is currently a Professor and Ph.D. supervisor. From March 2011 to September 2011, he
was an Advanced Visiting Scholar in Stevens Institute of Technology,
Hoboken, USA. His current research interests span the areas of wireless communications and signal processing, with emphasis on system
design of software defined radio, cognitive radio, and spectrum management.

Guoru Ding received his B.S. degree (with honors) in electrical engineering from Xidian University, Xian, China, in 2008. He is currently
pursuing his Ph.D. degree in communications and information systems
in College of Communications Engineering, PLA University of Science and Technology. His research interests include cognitive radio networks, machine learning, statistical signal processing, and big data analytics over wireless networks. He currently serves as a TPC member
of IEEE GLOBECOM 2014 and IEEE VTC 2014-Fall and an invited
reviewer for 10+ Journals, such as IEEE Signal Processing Magazine,
IEEE Communications Magazine, IEEE Transactions on Communications, and IEEE Transactions on Wireless Communications, etc. He
was a recipient of the Best Paper Award from IEEE WCSP 2009. He
is a IEEE/ACM student member and was a voting member of IEEE
1900.7 White Space Radio Working Group.

123