On the Essence of Knowledge-Aided Clutter Covariance Estimate and Its Convergence

Space-time adaptive processing (STAP) is a popularly used strategy for clutter suppression in moving-platform radar. In this context, the estimate of the clutter covariance matrix (CCM) is usually required to derive a near-optimum processing. The problem of estimation convergence then arises, especially in heterogeneous clutter environments, where, in the case of low convergence, the limited number of training samples will result in significantly degraded performance. Recently proposed knowledge-aided (KA) approaches show strong capability in improving convergence. Such capability is shown here to be essentially due to the reduction on the number of degrees of freedom (NDoF) of the sample space of the clutter process that bounds the convergence. In addition, the convergence measure of effectiveness (MOE) of two primary KA approaches, i.e., colored loading (CL) and fast maximum likelihood with assumed clutter covariance (FMLACC), is theoretically analyzed. The application of covariance matrix tapers (CMT) is proposed to enhance their robustness against knowledge mismatches. The simulation verifies the conclusions.