Hidden Markov Models for digital modulation classification in unknown ISI channels

This paper addresses the problem of classifying digital linear modulations transmitted through an unknown finite memory channel. Hidden Markov Models (HMMs) are used to model the received communication signals. In a classification purpose, our main interest is to determine the posterior probabilities of these received signals conditionally to each class. This paper proposes to use the Baum-Welch algorithm to compute these probabilities which are then plugged into the optimal Bayes decision rule. The performance of the proposed classifier is assessed through several simulation results.