基于多域特征的多雷达信号自动识别方法

In complex electromagnetic environments, simultaneous signal transmission of multiple radars in similar frequency bands can cause overlapping of reconnaissance's received signals in the time and frequency domains. To address this problem, an automatic modulation recognition (MR) method with a low signal-to-noise ratio (SNR) is proposed. The energy of the received signal is concentrated on the respective Doppler frequency through the row FFT transformation between the received pulses to achieve the separation of multiple intercepted signals. The Wigner-Ville distribution (WVD) and ambiguity function (AF) of each signal are simultaneously sent to the trained residual neural network (ResNet) as two layers of an image to solve the problem of low recognition accuracy when some types of signals only use a single time-frequency distribution. Theoretical analysis and simulation results show that the proposed method can not only effectively separate multiple radar signals overlapping in the time and frequency domains, but also accurately identify their modulation modes with the SNR of −10 dB.