Riemannian Mean Shift-Based Data Fusion Scheme for Multi-Antenna Cooperative Spectrum Sensing

In this article, the multi-antenna cooperative spectrum sensing problem in cognitive radio networks is investigated over Riemannian manifold. At the beginning, a signal matrix is constructed by using the sensing signals from secondary users (SUs) and the corresponding covariance matrix is calculated. Subsequently, the covariance matrices are transmitted to the fusion center and mapped to points on the manifold. In order to reduce the impact of aberrant SUs, a data fusion scheme based on Riemannian mean shift algorithm is developed. After data fusion, the representative points are obtained to train a classifier. In order to realize clustering directly over Riemannian manifold, a novel Riemannian distance based particle swarm optimization (RDPSO) algorithm, is proposed to train a classifier, which is employed to determine the state of primary user (PU). Finally, in simulation part, the validity of the proposed scheme is verified under different scenarios.

I. Introduction

With the continuous development of wireless communication network, spectrum resource is becoming increasingly scarcity [1]–[3]. However, most of the spectrum resources are not using sufficient [4]. In order to alleviate the shortage of spectrum resources, cognitive radio networks (CRN) have emerged [5], [6]. Different from traditional methods, CRN introduces cognitive radio (CR) technology to enable secondary users (SUs) to perceive the status of authorized channel. The SU can opportunistically access the available channel for communication when the primary user (PU) is not on the channel [7]. There are several spectrum sensing algorithms, such as energy detection (ED) [8], [9], matched filtering detection (MFD) [10], [11] and cyclostationary feature detection (CFD) [12], [13] have been proposed. However, these methods are restricted in practical application due to some shortcomings, such as require the prior information of the PU, susceptible to noise uncertainty and require a large amount of computation.