On detection of nonstationarity in radar signal processing

Space-time adaptive processing (STAP) has become a leading technique in airborne radar signal processing. The optimality of the STAP assumes the stationarity of the covariance matrices. In practice, however, the covariance matrices may be nonstationary. If such nonstationarity is not detected and not well treated, the STAP system's performance decreases substantially. In this paper, we present two detectors for detecting the covariance matrix nonstationarity. We form the first detector based on generalized likelihood ratio test, which inherits the property of asymptotically optimal detection performance. A second detector employs Rao test and requires significantly less computation than the first detector, which can be the favorable choice when computation load is of concern to the signal processing system.