A Novel Training Samples Selection Method for Space-Time Adaptive Processing

Abstract:
Estimating the clutter covariance matrix using the training samples is the key step of space-time adaptive processing (STAP). However, a lot of target-free training samples are not available in heterogeneous environments, resulting in the inaccuracy of the covariance matrix. Aiming at the problem, we propose a method using Pseudo-spectral as the test statistic to select training samples. The training sample whose Pseudo-spectral is similar to the Capon spectrum is selected. Moreover, we use subaperture smoothing techniques to estimate covariance matrix which can overcome samples deficiency problem. Finally, the numerical experiments with simulation data and measured data demonstrate the performance of the proposed method.