Cooperative Cumulants-Based Modulation Classification in Distributed Networks

Abstract

Automatic modulation classification of unknown wireless signals has been studied as a key technology in several cognitive radio and military applications. However, few studies have examined the performance of cooperative classification, in which spatially scattered nodes share their data to generate an overall decision of the unknown modulation scheme. In this paper, we propose a centralized feature-level cooperative classification framework using maximum likelihood (ML) combining algorithm. Each node estimates a set of features, based on high-order cumulants, that are combined at a fusion center using ML to produce a final decision. We present new theoretical models for ML combining performance, as well as computer simulations, under both additive white Gaussian noise and flat Rayleigh fading channels. In addition, a new hierarchical framework for cooperative modulation classification of larger sets of modulation schemes is presented. Finally, the performance of the cooperative framework with correlated signals is studied. Analysis and simulations validate the theoretical models and prove the efficiency of the proposed cooperative framework under different scenarios.