Download as pdf or txt
Download as pdf or txt
You are on page 1of 4

Spectrum Sensing Using Collaborative Beamforming for Ad Hoc Cognitive Radio Networks

Jing Zhang, Zheng Zhou, Haipeng Yao, Lei Shi, Liang Tang, Zhigang Xu Key Lab of Universal Wireless Communications, MOE Wireless Network Lab, Beijing University of Posts and Telecommunications Beijing, P.R.China E-mail: zjguet@gmail.com, zzhou@bupt.edu.cn

AbstractSpectrum sensing is one of the most important technologies for cognitive radio. In order to decrease the effect of low SNR, and increase the reliability spectrum sensing performance, we propose a novel spectrum sensing approach, which collects the distributed collaborative beamforming signals as the test statistic to input the energy detector. In particular, the collaborative beamforming method can enhance the received SNR in the desired direction of mainbeam, and reduce the input noise power. The distributed second users construct Ad Hoc network by self-organized form, making use of collaborative beamforming scan space signal, and can provide the general direction information for potential primary user. The numerical simulation results show that the potential improvement in primary user detection performance. Keywords-Cognitive Radio; Spectrum Sensing; Collaborative Beamforming; Energy Detection

Therefore, the SNR is a very important parameter for the primary user detection, especially for the hidden node problem which could result in the low SNR. As a useful tool, multiple antenna has been used in signal detection and energy-efficient signal transmission [2] [3], such as MIMO and smart antenna technology, these are centralized control manner. Moreover, aimed at distributed network, multiple users collaborative beamforming, which is used in wireless sensor Ad Hoc network to transmit data more effectively [4]. Each user equipped with a single isotropic antenna in this distributed network. Since, CRN network environment is changing more frequently, this distributed collaborative beamforming method even more to adapt the dynamic network environment. Motivated by the need of achieving reliable detection, it may possible to beamformer when received data in a distributed collaborative manner. Therefore, we propose an approach that a set of second users form Ad Hoc CRN, by collaborative beamforming to improve the detection performance in low SNR regions. The second users can be viewed as a virtual antenna array to imitate multiple antenna behavior in distributed CRN. The rest of this paper is organized as follows. Section II gives a brief description aiming at the distributed collaborative beamforming in CRN. Section III introduces a low-complexity and simple delay-and-sum beamformer algorithm for distributed collaborative beamforming. In section IV, we analyze the performance of single user and multiple users collaborative beamforming spectrum sensing with energy detector. In section V, our numerical simulation results are shown. In section VI, some conclusions are drawn. II. SYSTEM MODEL

I.

INTRODUCTION

Spectrum scarcity is becoming a serious problem in wireless communication systems. However, Federal Communications Commission (FCC)s report shows that the reason is low spectrum utilization. The cognitive radio concept, which was first proposed by Mitola in 1999, is a promising approach to achieve dynamic spectrum sharing [1]. As a generic approach, cognitive radio networks (CRN) exploit cognitive in wireless networks: using cognitive principles to improve utilization of resource. In CRN, the users can generally be divided into two categories, i.e., primary users and second users. The primary users (spectrum licensee) which have the higher priority to the spectrum resource are willing to share resources with second users. The second users are aware of the communication environment, intelligently detecting primary users behaviors, and determine when and how access the unused resources. The second users can dynamically adjust their radio operating parameters to modify system operation. The spectrum sensing technique plays a very important role in CRN. The second users sense spectrum environment to find an unused spectrum hole as quickly and accurately as possible. It notes that, in order to detect primary user, the receiver of second user must be more sensitive than the primary users.
This work was supported by NSFC (60772021), National 863 Program (2009AA01Z262), The Research Fund for the Doctoral Program of Higher Education (20070013029), Important National Science & Technology Specific Projects (2009ZX03006-006/-009) and Korean Ministry of Knowledge Economy Project (IITA-2009-C1090-0902-0019).

In this section, as illustrated in Fig.1, we describe a CRN consisting of M second users and a primary user located at the mainbeam direction [4]. We assume idealized antenna properties. Without loss of generality, the second users form an Ad Hoc network, and the geometry of the cluster is twodimensional. To join the CRN, a secondary user, as cluster head, will broadcast the hello message to the nearby secondary users. Thus, these related secondary users form an Ad Hoc network, and the geographical position information will be converged to the cluster head, which will perform spectrum sensing through collaborative beamforming. The spectrum sensing result would be share with the collaborative users.

978-1-4244-7010-5/10/$26.00 c2010 IEEE

1196

ISCIT 2010

Fig. 2. The delay-and-sum beamformer algorithm Fig. 1. Collaborative beamforming spectrum sensing model

The major difference between classical multiple antenna beamforming and distributed collaborative beamforming is that the geometrical information of the former is usually known as a priori, while the exact location of the second user nodes in Ad Hoc networks is unknown, and it should be acquired dynamically. Furthermore, for classical multiple antenna beamforming, the synchronization can be achieved in a centralized manner with reference to a RF carrier signal supplied by a local oscillator. But in the distributed network, each user has its own oscillator, the synchronization and geographical position and direction information among second users is critical, which may be achieved by the use of reference signals, such as those of Global Positioning System (GPS) [5]. III. DISTRIBUTED COLLBORATIVE BEAMFORMING

As mentioned above, the cluster head can obtain position and direction information among second users by GPS. The phase difference a ( i ) has relations with mainbeam azimuth i . Thus, the cluster adjust weights vector w(i ) is

w(i ) = [ w1 (i ), w2 (i ),! , wM (i )]T =

[ a1 (i ), a2 (i ),! , aM (i )]
M y (t ) = w(i ) H s(t ) .

(3) .

The output of the beamforming can be expressed as follows (4)

Therefore, the cluster output average power is P = E[ y (t ) y (t )] = w(i ) H R w(i ) , where R is the correlation matrix define by
R = E[ s (t ) s H (t )] E s12 (t ) E [ s1 (t ) s2 (t ) ] E [ s2 (t ) s1 (t )] E s2 2 (t ) = # # E [ sM (t ) s1 (t ) ] E [ sM (t ) s2 (t ) ] " E [ s1 (t ) sM (t ) ] . (6) " # % # 2 " E sM (t )

(5)

Multiple antenna beamforming has been applied in many fields, especially in MIMO system and smart antenna system. The MIMO system utilizes the space signals independence to achieve diversity gain, while the smart antenna system utilizes the correlation received space signal to obtain the SNR gain. We adopt the signal correlation theory to steer beampattern to the desired direction. Many beamformer algorithms are studied [6], such as delay-and-sum beamformer, adaptive Null Steering, etc. Delay-and-sum beamformer algorithm is a low-complexity algorithm for Space-Time processing, which is performed through coherent reception of amplitude and phase of the signal received from each antenna elements array [7]. Here, we adopt delay-and-sum beamformer algorithm for the distributed collaborative beamforming. As shown in Fig.1, suppose that the primary users signal is the narrow band far field signals. The ith user received signal si (t ) can be expressed as

It is noted that since the signals received at the traditional multiple antenna are complete correlated, the SNR gain can reaches M times [6], for the distributed collaborative beamforming algorithm, which the correlation of the received signal is weakened, the SNR gain can not reach M times. The reason is that the different user received signals pass different path, thus the path loss is different in distributed network. We assume the correlation matrix R with the element Rii = 1 , and the other element follows uniform distribution Rij (i j ) [0,1] . Thus, the matrix R is determined by the non main diagonal element. In order to simplify the analysis, we denote by Ps the average power of received signal for each second user. Based on (6), the cluster output average power P can be presented as

si (t ) = m(t ) exp[ j (2 f c t + i )] ,

(1)

where m(t ) is the complex modulating function of received signal, f c is carrier frequency, i is the phase . Thus, the input cluster of signal vector for all second users can be expressed as
s (t ) = [ s1 (t ), s2 (t ),! , sM (t )]T . (2)

We consider the path loss being with different phase, the noise is n i (t ) , Additive White Gaussian Noise (AWGN). Thus, as depicted in Fig.2, the weights are applied here to adjust phase for the maximize SNR. The delay-and-sum beamformer algorithm can be further depicted as follows

P = Ps

R
i =1 j =1

ij

(7)

2 In a similar way, aiming at the zero mean and variance n 2 AWGN, the noise is uncorrelated, R = n I . Thus, the output

1197

noise average power is M times less than on each user, it can be expressed as follows
Pn = E[n(t )n (t )] = w(i ) H R w(i ) =
2 n

Thus, the probability of detection P d and the probability of false alarm Pf for AWGN are given by

(8)

Thus, we define the cluster output average SNR gain for distributed collaborative beamforming, which is the ratio of the cluster output SNR to the input SNR, equal to M the number of users in the cluster. When the primary signal locates in the mainbeam region, the algorithm can provide about ( Rij ) / M times SNR gain.
i =1 j =1 M M

(13) Pf = P[Y > | H 0 ] = (TsW , / 2) / (TsW ), where () and (, ) are the complete and incomplete gamma functions, is detection threshold, and Qm (.,.) is the generalized Maccum Q-function defined as follows Qm (a, b) =
+

Pd = P[Y > | H1 ] = Q2TsW ( 2 , )

xm x e a m 1

+ a2 2

I m 1 (ax)dx ,

(14)

IV.

THE SPECTRUM SENSING SINGLE USER WITH ENERGY DETECTOR

where I m 1 (.) is the modified Bessel function of (m 1)th order. B. The detection performance of multiple users collaborative beamforming In this subsection, we will analyze the performance of collaborative beamforming. On the basis of the abovementioned results in Section II, the received signal can be presented as
w(i ) H [h1 x1 (t ) h2 x2 (t )! hM xM (t )]T + n(t ) r (t ) = n(t )

A. The detection performance for single user In traditional single user signal detection, the goal is to determine whether the signal is present (hypothesis H1 ) or not (hypothesis H 0 ). Under these two hypotheses, the received signal is given by [8]

hs (t ) + n(t ) r (t ) = n(t )

H1 H0

(9)

H1 , (15) H0

Suppose that N 0 is the two-side noise power spectrum density (psd). s (t ) is the primary user signals. h is channel gain between primary user and second user. We consider a modified energy detector at different hypotheses based on the normalized quantity, Y = Yr / N 0 , where Yr is the energy of the received signals under the two hypotheses. Under H 0 , the output of the energy detector can be express as follows 1 Y= N0

with the average SNR on the ith user. The cluster output average SNR CB is given by

CB =

R
i =1 j =1

ij

(16)

Ts

1 n (t )dt = 2 N 0W
2

2TsW i =1

2 i

(10)

Under hypothesis H1 , Y has a non-central chi-squared distribution with 2TsW degrees of freedom and the noncentrality parameter 2( Rij / M ) = 2 CB
i =1 j =1 M M

thus

the

where the signal duration time is Ts , the bandwidth is W , Based on the Nyquist sampling theorem, the number of the sample points is denoted by 2TsW . Over the interval (0, Ts ) , we have
n (t ) =
2TsW i =1

probability of detection P d is presented as Pd = P[Y > | H1 ] = Q2TsW ( 2 CB , ) . (17)

n sin c[(2Wt i)] ,


i

(11)

Under hypothesis H 0 , based (10) the energy detector output can be express as Y= 1 2 N 0W
2TsW i =1

where sin c( x) = sin( x) / x , ni = n(i / 2W ) . Since, ni (t ) is i.i.d and ni ~ N (0, 2 N 0W ) , under H 0 , Y has a central chi-squared distribution with 2TsW degrees of freedom. In a similarly way, under H1 , Y has a non-central chisquared distribution with 2TsW degrees of freedom and the non parameter 2 , where is the SNR.

2 i

1 2 N 0W i.i.d

2TsW i =1

( w (i) n ( 2W ))
k k k =1

. (18) with

Where

nk (t )

is

for

all

second

users

2 distribution N (0, n ) . Consequently, the random variable follows distribution

H1 : Y ~

2 2TsW 2 2TsW

(2 )

M k =1

2 wk (i ) nk (i / 2W ) ~ N (0, n 2 ) , M 2

(19)

H0 : Y ~

(12)

2 where 2 = n k =1 wk (i )

. The Probability Distribution

Function (PDF) of Y under hypothesis H 0 is given by [8]

1198

fY ( z ) =

1 z z TsW 1 exp( 2 ) . 2 (2 ) (TsW )


2 TsW

(20)

Thus, the Cumulative Distribution Function (CDF) Y is FY ( z ) = fY ( z )dz = 1


0 z

(TsW , z /(2 2 )) . (TsW )

(21)

The probability of false alarm Pf for the AWGN channel is presented as Pf = P[Y > | H 0 ] = V. (TsW , /(2 2 )) . (TsW ) (22)

SIMULATIONS AND ANALYSES

Fig. 4. The probability of detection with different collaborative users (C=0.6)

We consider a distributed Ad Hoc CRN in which the second users find unused frequency hole to access. Assume the secondary users are located following the uniform distribution. To analyze the sensing performance of the proposed method, we investigate the detection probability P d for the fixed false alarm probability Pf at given SNR in AWGN channel. The SNR is varied from 0dB to 20dB, Pf = 0.001 , TsW = 6 . Base on (7), we assume the average correlation factor C = Rij (i j ) . First, we analyzed the detection performance with different correlation factor. Let M = 6 , C = 0.2, 0.4, 0.6,1 , respectively. As shown in Fig.3, with the increasing of the correlation factor, the probability of detection P d is significantly improved. The ideal condition is C = 1 . The idea behind this is that when the signal is perfect correlated, the collaborative beamforming method can reach the maximum SNR gain. But in practice, considering the second users distributed manner, the received signal through different channel could not be complete correlation.

VI.

CONCLUSION

In this paper, we proposed a second users collaborative beamforming method, which can increase the received SNR to improve the energy detector performance for CRN spectrum sensing comparing to the single user energy sensing scheme. The main idea behind this is that second users who can selforganize Ad Hoc CRN could detect primary users signal in the direction of mainbeam meanwhile reduce the input noise for energy detector. Moreover, the delay-and-sum beam former algorithm holds the characteristics of lower computing complexity, and need no channel information. Thus, it is fit for distributed network environment. It is worth noting that the different users received signal should keep a certain correlation for collaborative beamforming. Consequently, the geometrical configuration and synchronization of distributed users have a great impact on detection performance, and it is the future research aspect. REFERENCES
[1]

[2] [3]

[4]

[5]

[6] Fig. 3. The probability of detection with different correlation factor (M=6)

As shown in Fig.4, we examine the spectrum sensing performance with different users. In this case, let C = 0.6 , M = 2, 4, 6 , respectively. The results show that the probability of detection P d is improved with the increasing of collaborative users.

[7] [8]

Mitola, J, Maguire, G.Q, Jr, Cognitive radio: making software radios more personal, IEEE Wireless Communications, 1999, Vol.6, No. 4, pp.13-18. A. Molisch, wireless communication, Wiley-IEEE Press, 2005. Xiaolong Ma, Guican Chen, Robust adaptive beamforming for cyclostationary signals in spectrum detection of cognitive radios. International Conference on Networks Security, Wireless Communications and Trusted Computing, 2009, Vol.1, pp.387-390. Yu Deng, Burr A.G, Pearce D, Grace D, Distributed beamforming for cognitive radio networks, ChinaCom, 2008, pp.1206-1210. Ochiai.H, Mitran.P, Poor, H.V, Tarokh. V, Collaborative Beamforming for Distributed Wireless Ad Hoc Sensor Networks, IEEE Transactions on Signal Processing, Vol.53, pp.4110-4124. N. M. Neihart, S. Roy, and D. J. Allstot, A parallel, multiresolution sensing technique for multiple antenna cognitive radios, IEEE International Symposium on Circuits and Systems, 2007, pp.2530-2533. Hult T, Mohammed A, Cooperative Beamforming for Wireless Sensor Networks, EuCAP 2007, pp.1-4. A. Pandharipande and J.-P. M. G. Linnartz, Performance analysis of primary user detection in a mutiple antenna cognitive radio, IEEE International Conference on Communications, 2007, pp.6482-6486.

1199

You might also like