SINR Rank Ordering Metric for signal dependent sub-optimum STAP

In a previous work the authors have developed an algorithm, based on the method of Principal Components, that exhibits the advantage of further reducing the training samples necessary to obtain acceptable SINR losses for the estimation of the space-time covariance matrix. The new algorithm subject of the previous paper is named Reduced Dimension Principal Components (RDPC) STAP. The sub optimum ADPCA STAP uses the inverse of a space-time sub matrix associated to set of P pulses to build the STAP filter. The idea under RDPC is that the ADPCA inverse sub matrix may be approximated using the method of Principal Components (PC) that is retaining the clutter (dominant) eigenvectors/eigenvalues pairs to build an effective covariance inverse. In the RDPC paper the authors have shown that clutter exhibits a rank K that is smaller than the dimension N by P of the sub matrix thus allowing a number of approximately 2K independent training cells to estimate the space-time sub matrix thus achieving about 3 dB losses on the SINR after the STAP filter. In this paper the authors propose a method for selecting the clutter eigenvectors/eigenvalues pairs of the space-time sub matrix used to build the STAP filter. The proposed method is based on a SINR Rank Ordering Metric (ROM) defined over the sub matrix, thus including signal dependence in the algorithm. Simulations show good results in terms of SINR losses with respect to optimum STAP also considering the case of range cell migration (RCM).