Maximum likelihood sequence estimation via the expectation maximization algorithm in the presence of random phase and amplitude fading

Abstract

The expectation-maximization (EM) algorithm was first introduced in the statistics literature as an iterative procedure that under some conditions produces maximum likelihood (ML) parameter estimates. Here we investigate the use of the EM algorithm to the problem of estimating transmitted sequences corrupted with random phase and amplitude fading as well as additive white Gaussian noise. The EM-based algorithm is derived and is shown to have linearly growing complexity as a function of sequence length. The algorithm is applied to the random phase and fading channels and its performance is seen to approach that of the optimal ML sequence estimator, which however is too complicated to implement in practice.