STAP detection using space-time autoregressive filtering

Application of space-time adaptive processing (STAP) in real situations requires dimension-reducing methods. This is due to both the large computational cost involved in calculating the interference statistics and the smaller number of stationary training samples available to estimate the clutter covariance. Recently, auto-regressive (AR) filtering techniques have been used to help reduce computation and secondary sample support requirements in STAP scenarios. We compare the detection performance of several AR-based algorithms with more standard GLRT-type approaches. In particular, we consider the parametric amplitude matched filter (PAMF) and the space-time autoregressive filter (STAR), and show that they outperform standard GLR tests, especially in challenging situations with low sample support. Among the parametric methods considered, the STAR approach provides the most robust overall performance.