A general maximum likelihood classifier for modulation classification

This paper deals with maximum likelihood (ML) classification of digital communication signals. We first propose a new approximation of the average likelihood function. Then we introduce the General Maximum Likelihood Classifier (GMLC) based on this approximation which can be applied to a wide range of classification problem involving finite mean power signals. Derivation of this classifier equations are given in the case of linear modulations with an application to the MPSK / M PSK problem. We show that the new tests are a generalization of the previous ones using ML approach, and don't need any restriction on the baseband pulse. Moreover, the GMLC provides a theoretical foundation for many empirical classification systems including those of that exploit cyclostationarity property of digital modulated signals.