Abstract:

Performance of the sample covariance matrix, which is estimated with training samples contaminated by targetlike signals, decreases in space time adaptive processing (STAP). To mitigate the problem, a novel training sample selection algorithm based on sparse reconstruction is proposed. This algorithm firstly transforms the domain of the received data from array elementpulsedistance to array elementDopplerdistance, then recover the treated data to obtain the high resolution angledistance spectrum using spatial sparse reconstruction based on the refined FOCal underdetermined system solver (FOCUSS). Later, find and discard the samples of the measured angle, which significantly differs from the expected angle of the angledistance spectrum, with the prior knowledge of clutter Doppler and angle of incidence, and finally make an efficient choice of the training sample. The theory analysis and experimental results illustrate that compared with the traditional method, with small sample set the proposed method screens out the contaminated training sample effectively and improves the performance of STAP without estimating the sample covariance matrix.