Semantic Graph Reasoning over Multi-Domain Fusion for Radar Jamming Recognition

Radar jamming identification is a key technology in radar anti-jamming systems, providing a foundation for intelligent countermeasure decision-making. To address the limitations in multi domain feature fusion of radar jamming signals, this paper proposes a method based on Multi-Domain Fusion for Semantic Graph Reasoning (SGR-MDF). A dual-domain feature interaction network is designed to enable adaptive complementarity and joint enhancement of time- and frequency-domain features. Multi-level features from the time, frequency, and time-frequency domains are then fused to improve signal representation. A graph convolutional classifier is introduced to model both comprehensive and core semantic relationships among jamming types, enhancing classification performance. Experimental results on a dataset comprising 50 jamming types demonstrate that the proposed method achieves a recognition accuracy of 98.68% at a jamming-to-noise ratio (JNR) of 0 dB, outperforming state-of-the-art approaches and exhibiting strong robustness under low-JNR conditions.