Information geometry and estimation of Toeplitz covariance matrices

The estimation of covariance matrix is of fundamental importance in radar signal processing. Recent work has shown that information geometry provides a novel approach to estimating the covariance matrix. Prior work has shown that an information geometry inspired covariance matrix estimator provides significant gains (in SINR loss terms) over several standard estimators, such as the loaded sample matrix inversion (LSMI). In this paper, some techniques for computing the covariance matrix, inspired by information geometry, are presented. It is found that some algorithms provide superior performance when the number of samples is small.