Performance analysis of data sample reduction techniques for STAP

To detect and identify targets in changing interference environment that includes clutter and jammers, space time adaptive processing (STAP) algorithms can be utilized. Often in nonstationary clutter, the available stationary sample support data is severely limited to be useful for direct implementation of the sample space-time covariance matrix inversion approach for optimal detection. In this paper we outline and compare two new approaches to address the sample support problem: (i) generalized forward-backward sub-aperture-subspace smoothing techniques to reduce the number of data samples in estimating the sample covariance matrices; and (ii) projection methods using alternating projections or relaxed projection operators onto desired convex sets to retain the a-priori known structure of the covariance matrix. Performance comparisons are presented to show that by utilizing these approaches with eigen based techniques, it is possible to reduce significantly the data samples required in non-stationary environment and consequently achieve superior target detection.