DS-Diff-SwinT: Diffusion model and D-S embedded swin transformer integrated framework for radar jamming recognition

Radar anti-jamming is essential for the secure operation of radar systems in hostile environments, and accurate recognition of radar jamming types is critical. With increasing environmental noise and more diverse jamming patterns, recognizing radar jamming becomes increasingly challenging. To address this problem, this paper presents a diffusion model and Dempster-Shafer (D-S) embedded Swin Transformer integrated framework (DS-Diff-SwinT). This approach involves generating a time-frequency distribution map from jamming signals using the Choi-Williams distribution (CWD) and subsequently utilizing the swin transformer to capture multi-scale features within different receptive fields. Finally, a D-S fusion algorithm is embedded within the model to integrate recognition confidence across these fields, yielding the final recognition outcome. Experimental results on eight radar jamming categories demonstrate that this method outperforms traditional approaches in recognition performance.