Stochastic maximum likelihood methods for semi-blind channel estimation

A blind stochastic maximum likelihood (ML) channel estimation algorithm is adapted to incorporate a known training sequence as part of the transmitted frame. A hidden Markov model (HMM) formulation of the problem is introduced, and the Baum-Welch (1970) algorithm is modified to provide a computationally efficient solution to the resulting optimization problem. The proposed method provides a unified framework for semi-blind channel estimation, which exploits information from both the training and the blind part of the received data record. The performance of the ML estimator is studied, based on the evaluation of Cramer-Rao bounds (CRBs). Finally, some preliminary simulation results are presented.