Performance of knowledge aided space time adaptive processing

In this study, the asymptotic average signal-to-interference plus noise ratio (SINR) loss of knowledge-aided (KA) space time adaptive processing (STAP) is derived based on the random matrix theory. The authors observe that the desired steering vector and a priori covariance matrix is whitened by the covariance matrix of cell under test. An important result in this study is that one finds the SINR loss of KA STAP can be factorised into two parts. The first part of SINR loss is determined by the number of independent and identically distributed secondary samples, system degree of freedom, colour loading level and the eigenvalues of whitened a priori covariance matrix. The angle between two vectors accounts for the second part of SINR loss, where the first vector is the whitened desired steering vector, the second vector is a rotated version of this whitened vector with the rotation matrix equal to whitened a priori covariance matrix. Several cases including both accurate a priori knowledge and inaccurate a priori knowledge as the colour loading matrix are studied. Performance loss because of inaccurate a priori knowledge is analysed. Numerical simulations are conducted to validate the authors theoretical analysis, shows that the results based on random matrix theory fit the empirical results by Monte-Carlo trials well.