An improved ResNet-ViT Fusion Model for Satellite Communication Jamming Recognition

In non-cooperative reception scenarios, satellite communication signals are susceptible to the effects of intentional jamming, especially suppression jamming. Correctly detecting and identifying jamming is crucial for enhancing the antijamming capability of the system. However, the time-frequency overlap between jamming signals and communication signals significantly increases the difficulty of recognition. To address this issue, this paper proposes a sequence-input fusion neural network classification model based on ResNet and Transformer, which enables the identification of target communication and jamming signals. By jointly learning time-domain and frequency-domain features, the proposed model significantly improves jamming detection and recognition accuracy. Experimental results demonstrate that when the jamming-to-signal ratio (JSR) is below 4 dB, the fusion model outperforms the pure ResNet model in extracting features of frequency-modulated jamming and achieves a significant improvement in overall recognition accuracy. This validates the effectiveness of the proposed method in complex electromagnetic environments, providing new technical support for optimizing satellite communication anti-jamming methods and strategy transformations.