Signal Estimation based on Mutual Information Maximization

We study the problem of estimating signals in the presence of noise, clutter, and interference based on maximization of mutual information. Traditional approaches for signal estimation involve minimum mean squared error, maximum likelihood, minimum variance unbiased estimators, and others. The derivation of estimators for most of these, however, requires precise knowledge of the signal and noise (or clutter and interference) distributions. In many practical applications, these are difficult to obtain. Here we propose a generic approach for estimating signal parameters by maximizing the mutual information between the signal being estimated and the available data. We show by simulation that this approach can significantly outperform least squares-based approaches in estimating parameters of linear models, including an application in time-delay estimation.