Bayesian multi-class covariance matrix filtering for adaptive environment learning

Abstract

Covariance matrix estimation is a crucial task in adaptive signal processing applied to several surveillance systems, including radar and sonar. In this paper we propose a dynamic environment learning strategy to track both the covariance matrix and its class; the class represents a set of structured covariance matrices. We assume that the posterior distribution of the covariance given the class, is basically a mixture of inverse Wishart, while the class posterior distribution evolves according to a Markov chain. The proposed multi-class inverse Wishart mixture filter is shown to outperform the class-clairvoyant maximum likelihood estimator in terms of covariance estimate accuracy, as well as the Bayesian information criterion rule in terms of classification performance.