Robust covariance matrix estimation for space-time adaptive processing

Clutter covariance matrix estimation in nonhomogeneous environments is one of the main concerns in Space-Time Adaptive Processing (STAP). A number of estimation methods have been proposed by exploring special structure of the covariance matrix. However, these methods are often difficult to be implemented in practical situations. In this paper, we propose a new method. In this method, the conventional estimated covariance matrix is first split into clutter and noise part through eigendecomposition, a structured form of noise part is then estimated. With the new noise part and the original clutter part forms the new covariance matrix. Finally, the beamforming is formed based on the new covariance matrix. Computer simulations demonstrate that STAP based on the new approximation is robust both in very limited sample support case and in gain and phase errors exist case.