Cooperative Combining of Cumulants-Based Modulation Classification in CR Networks

Abstract

Automatic Modulation Classification (AMC) is a key enabling technology in Cognitive Radio (CR) Networks. The ability of CR transceivers to detect and classify unknown wireless signals has various applications in civilian and military domains. Performance of AMC degrades severely under low Signal-to-Noise Ratio (SNR) and variable channel conditions. Cooperative classification has been presented as a means to overcome the detrimental channel effects by combining the results from physically scattered CR nodes. In this work, Maximum Likelihood (ML) combining of classification features is presented as a data fusion algorithm that provides better classification accuracy compared to hard decision combining algorithms without high network overhead. The performance of a cumulants-based modulation classifier under Additive White Gaussian Noise (AWGN) is analyzed. The enhancement in classification performance when applying ML combining of more than one classifier is presented. Theoretical analysis as well as various simulations are presented for ML combining of CR nodes with equal SNR. In addition, analysis is extended to the case where CR nodes have different SNRs. Theory and simulations show that applying ML combining will result in a better classification accuracy, even when one of the nodes has a much lower SNR.