Denoised Jarque-Bera features-based K-Means algorithm for intelligent cooperative spectrum sensing

Abstract

This paper aims to boost the blind perception performance of a cooperative spectrum sensing system in a noisy environment. A novel intelligent centralized-cooperative spectrum sensing technique based on new denoised features developed by the Jarque-Bera (JB) statistics test and the K-Means (KMS) clustering algorithm is proposed (JBKMS). The skewness and access kurtosis are high-order moments of JB statistics that are utilized to build denoised features and attenuate the noise level of sensing samples from secondary users. The new denoised features are named and formed by four mathematical criteria, including the difference between large and average JB (DLAJB) vector values; the difference between large and small JB (DLSJB) vector values; the ratio between large and small JB (RLSJB) vector values; and the ratio between large and trace JB (RLTJB) vector values. In comparison to other standard strategies, the proposed methods eliminate the computational complexity associated with threshold estimation and improve sensing performance by minimizing interference across KMS clusters. Then, the KMS method is used to classify the denoised features based on the presence or absence of the primary user in the noise channel in order to construct a well-trained classifier. Various energy features of random matrix theory based on the clustering algorithm are compared in this work. The results of the simulations demonstrate that the devised methods have superior detection performance in terms of probability of detection at low signal-to-noise ratio.