Abstract: In the airborne radar signal processing, the strong ground clutter is a major problem affecting the signal detection performance. The space-time adaptive processing (STAP) is an effective technique to suppress the clutter. In practical processing, because of the non-stationarity of the clutter, the STAP usually faces the problem of a small number of available valid samples. In order to solve this problem, an angle-Doppler channel selection method based on sparse recovery is proposed. We utilize a small number of samples to estimate the full-dimensional clutter covariance matrix (CCM) via the sparse recovery method and evaluate the importance of each channel with the estimated full-dimensional CCM. Then we select the appropriate channels to construct the reduced-dimensional clutter covariance matrix for the STAP processing. The proposed algorithm can solve the problem of few samples with good performance of the reduced-dimension STAP (RD-STAP). The numerical simulation verifies that the algorithm is effective and better than several typical STAP algorithms. The relationship between output performance and the number of channels under different sample numbers is also discussed.