Polyphase-modulated radar signal recognition based on time-frequency amplitude and phase features
Xue Ni, Huali Wang, Yang Yu, Ying Zhu, Zhiguang Zhang · 2020
The current recognition method based on time-frequency analysis only uses the amplitude spectrum and ignores the phase spectrum, which leads to the recognition rate of radar signal low. In this paper, we propose an automatic recognition method based on time-frequency amplitude and phase features. For the signal time-frequency analysis, we take the short-time Fourier-based synchrosqueezing transform for radar signals to obtain two-dimensional time-frequency representation with complex values. Then, we extract the time-frequency amplitude spectrum and phase spectrum and take the absolute value of them as the two inputs of the recognition network. Next, we construct a deep convolutional network with two channels for automatic feature extraction and recognition. The simulation results on 5 kinds of polyphase codes show that the proposed method has superior performance in distinguishing the polyphase codes even at low SNR, and the phase features help to improve the recognition rate.