A Radar Signal Modulation Recognition Algorithm Based on Multi-branch Fusion Difference

The time-frequency images of radar modulation signals are similar at low signal-to-noise ratios, which will lead to lower recognition accuracy. This paper proposes an algorithm based on multi-branch fusion differences to identify radar modulation signals. The algorithm designs a multi-branch fusion differential module based on the attention mechanism and differential thinking, to extract time-frequency image fine-grained features and improve the Convolution Neural Network characterization ability. After that, a dual-branch differential network is proposed by combining vanilla convolution and multi-branch fusion difference module. The network can fuse global and detailed features of time-frequency images, which can achieve radar signal modulation recognition. The simulation results show that the recognition accuracy of the algorithm can reach 93.17% at 0dB. Therefore, the algorithm can obtain high recognition accuracy and has high recognition performance.