IMPROVED EMPIRICAL COVARIANCE ESTIMATION

Covariance estimation through a measurement space method can be made more robust by incorporating weights into the least squares solution process. When the measurements are of full rank, it can be shown that the calculation replicates that of a normalized sample covariance calculation. Furthermore, distribution sampling theory can then be used to estimate errors in these covariance estimates. Lastly, this estimator is applied successfully to realistic scenarios where the error distribution is non-Gaussian in Cartesian space.