Cooperative multiuser modulation classification in multipath channels via expectation-maximization

With the advent of cognitive radio (CR) and dynamic spectrum access techniques, where multiple signals may coexist within the same frequency band, multiuser modulation classification problem becomes a vital issue, which has not been sufficiently investigated. In this paper, we consider a cooperative multiuser modulation classification problem, in the presence of unknown multipath channels. A likelihood-based (LB) classifier using the expectation-maximization (EM) algorithm is proposed, which enables to find the maximum likelihood estimates (MLEs) iteratively. Numerical results show that the proposed algorithm achieves significant improvement on the classification performance with a small number of samples when compared to the conventional methods, which demonstrates its reliability and efficiency of identifying modulations of multiple users under the multipath scenarios.