Radar Compound Jamming Recognition Based on Image Segmentation and Fused Attention Residual Network.

With the increasing complexity of modern electromagnetic environments, radar systems are not only affected by single jamming signals but also by compound jamming, which consists of additive combinations of multiple jamming types. Compound jamming is difficult to recognize due to a wide array of diverse compound patterns. To address this issue, this study proposes a method for the segmentation and recognition of compound jamming signals. First, a jamming segmentation module based on image segmentation techniques is designed to segment the compound jamming in the time-frequency domain, which is obtained by short-time Fourier transform (STFT). Subsequently, an enhanced residual network (ResNet) incorporating a spatial-channel fused attention mechanism (SCFAM) is proposed to effectively capture multi-level features and recognize the segmented jamming signals. The experimental results demonstrate that the proposed method achieves a recognition accuracy of 98.60% for compound jamming, outperforming three classical approaches. Additionally, this method exhibits superior performance in recognizing untrained types of compound jamming, highlighting its robustness and generalization capability.