Node Selection Based Distributed Cooperative Compressive Spectrum Sensing
Xuekang Sun, Rikang Zho, Jincheng Zhao, Li Gao
School of Network Education, Beijing University of Posts and
Telecommunications,
Beijing, China 100876
e-mail: sunxuekang6395@126.com
Muyan Ma
School of Instrument Science and opto Electronics Engineering
Beijing Information Science & Technology University
Beijing, China 100092
ma my@bistu.edu.cn
Abstract—In the existing distributed cooperative spectrum
sensing scheme, the nodes that participate in the cooperative
operation are predetermined. However, the nodes in the deep
fading wireless environment usually corrupt the performance
of cooperative sensing. To deal with this problem, we propose a
node selection based distributed cooperative compressive
spectrum sensing (NS-DCS) algorithm for wideband cognitive
radio network (CRN). In this scheme, we first use Bayesian
network learning to remove the redundant nodes based on the
compressive data collected from the different cognitive radio
(CR) users in a fading environment. And then we employ a
proposed weighted average consensus-based distributed
cooperative compressive spectrum sensing algorithm with the
prior parameter exchange (PE-WDS) to identify the occupied
spectrum. The simulation results show that our proposed
scheme can improve the accuracy of wideband cooperative
spectrum sensing.
Keywords-distributed cooperative spectrum sensing, node
selection, wideband cognitive radio network
I. INTRODUCTION
Cognitive radio has become a promising technique to
solve the spectrum scarcity problem for supporting evolving
wireless services and applications. In cognitive radio (CR)
systems, the unlicensed users can utilize the licensed
frequencies while the primary user (PU) is not active.
Therefore, spectrum sensing is a key technique for cognitive
radio which can efficiently identify the unoccupied channels
or spectrum holes to enable cognitive radio (CR) users to
access the idle spectrum without the influence on the primary
users (PUs) [1]. In the existing broadband spectrum
perception approach, compressive sensing (CS) is an
important way to alleviate the hardware requirements of the
receiver, by which the sparse signal sampled at sub-Nyquist
rate can be recovered [2]. However, as an individual CR user
is quite susceptible to the shadowing and fading, especially
in a lower signal-to-noise ratio (SNR) wireless environment,
cooperative spectrum sensing (CSS) through coalition games
among CR users is an effective solution of this problem [3].
Based on the different network control strategies, centralized
and decentralized data fusion methods are developed. In our
paper, we focus on the distributed cognitive radio network in
which each CR alternatively senses the signals of primary
users, exchanges sensing messages with its one hop
neighboring CR users and uses them to estimate the PU
spectrum.
To deal with the complicated operation of the distributed
cooperative sensing, several data fusion schemes are
investigated to increase the sensing accuracy corresponding
to different sampling strategies and application environments.
Reference [4] shows an average consensus based model in
which each CR user exchanges the sensing messages with its
neighboring nodes and uses them to optimize the fusion
operation until the fused massage converges to the average
value. In [5], the author proposes a probabilistic graphical
model to represent the characteristics of the nodes and fuses
multi-prior information from neighboring CR users in
heterogeneous cognitive radio networks. It is clear that the
nodes that participate in the cooperative operation are
predetermined in above schemes. However, the nodes in a
low signal-to-noise ratio (SNR) wireless environment usually
corrupt the performance of cooperative sensing. Some
researchers use many methods, such as a greedy strategy and
Boolean quadratic programming to derive the active sensor
selection[6][7]. As above algorithms are easy to fall into the
local optimal situation, we employ Bayesian network
learning to remove the redundant nodes in the proposed NS-
DCS algorithm.
The rest of this paper is as follows: Section II presents a
system model. Section III introduces our proposed NS-DCS
algorithm. In Section IV, the simulation results are presented
to demonstrate its performance of the proposed algorithm.
The conclusion is discussed in section V.
II. SYSTEM MODEL
We consider a distributed wideband CRN which
includes L CR users and primary users. Each PU may
simultaneously transmit a message on all or several non-
overlapping licensed sub-bands in the fading environment.
Each CR user senses the sparse signal by Bayesian sampling
model [8], and then recovers the original signal by the
Relevance Vector Machine (RVM) algorithm iteratively.
Meanwhile, a weight
i
is obtained, which denotes the
channel state. After that, each CR user exchanges its
recovery vectors and weight mutually with its one hop
neighboring nodes, and then uses the proposed PE-WDS
algorithm to decide whether the bands are occupied or not.
Obviously, there is a huge need for the exchange
information. In addition, some cooperative nodes which are
affected by the random noise, shadowing and multi-path
fading, make the sensing performance degrade. Therefore,
653
2017 3rd IEEE International Conference on Computer and Communications
978-1-5090-6351-2/17/$31.00 ©2017 IEEE