Contents I. Introduction
Early Warning Radar (EWR) can acquire not only the metric information of radar targets such as distance, moving direction, azimuth angle and elevation angle, but also radar cross section (RCS) [1] measurements which contain abundant signatures of targets. Distinguishing warhead and decoys in ballistic missile defense system (BMDS) [2] is a big challenge. After the separation of warhead from the fuel tank and release of the decoys, the warhead has regular motion caused by spinning and conning, while the motion of decoys is usually irregular caused by tumbling. Consequently, their micromotion characteristics are significantly different, resulting in different modulation characteristics of the radar returns. The information contained in the time sequences of RCS amplitude reflects recognizable target characteristic such as size, shape, structure as well as the state of motion. As a result, RCS measurements are considered to be important radar target signatures for the analysis of target micro-motion. With the continuous development of modern signal processing, the RCS feature extraction techniques continue to be improved [3], [4]. Past experiments demonstrated that fully exploiting the information contained in RCS is important for target recognition in BMDS [5].
