Radar Signal Recognition Based on DAVG-GRN Network

The increasing complexity of the electromagnetic environment poses significant challenges to the accurate identification of low probability of intercept (LPI) radar signals, especially under low signal-to-noise ratio (SNR) conditions. To address this issue, this paper proposes a novel deep-learning based framework DAVG-GRN for end-to-end radar signal recognition under low SNR conditions which integrates a Dynamic Adaptive Visibility Graph (DAVG) module and a Residual Graph Neural Network (GRN) classifier. Specifically, the DAVG module adaptively determines the convolution kernel size based on the input signal characteristics through an adaptive convolution selection mechanism. It then employs dilated convolution to map the original time-domain signal into a graph structure, thereby preserving both local patterns and long-range dependencies within the radar signal. The GRN module adopts a multi-layer residual graph convolution structure to deepen the network while preserving low-level feature information. Furthermore, multi-scale graph pooling is used to extract global discriminative features, enhancing the model’s representation capability and classification performance. Experimental results show that the proposed DAVG-GRN outperforms a series of advanced deep learning methods under low SNR.