Adaptive waveform design based on LSSVM for moving target recognition in cognitive radar

The optimal waveform for extended target recognition is directly affected by the target impulse response, which is sensitive to the target aspect. Hence, the variation of target aspect needs to be considered when the target is moving. Aiming at this problem, a new framework of cognitive radar is proposed. It predicts the new aspect via least square support vector machines (LSSVM) by using the prior knowledge of target aspect, and then obtains the optimal waveform based on not only the updated prior probabilities of the target hypothesis but also the updated TIR in a circular of interrogation. Simulations part shows the loss of recognition efficiency for a moving target when treated as static by the method in previous literature, and proves the validity of the proposed method.