Regularisation methods for covariance matrix estimation in low sample support stap

The interference covariance matrix in space time adaptive processing is usually estimated from training data that is drawn from adjacent range gates to the cell under test. The sample support required to achieve good detection performance may not be available in practice due to many factors such as system design or clutter heterogeneity. In this paper, we address this problem by proposing and evaluating a number of covariance matrix regularisation methods that attempt to reduce the bias of the eigenvalues of the estimated covariance matrix and hence improve the signal detection performance.