A Hybrid 1D-2D CNN with Decision-Level Fusion for Radar Active Deception Jamming Recognition

Radar active deception jamming patterns exhibit high versatility and dynamic characteristics in modern electronic warfare. In recent years, convolutional neural networks (CNNs) have become effective tools for signal processing tasks. In this study, we propose a decision-level fusion model that integrates time-domain features extracted by a 1D-CNN and time-frequency features derived from a 2D-CNN for radar jamming recognition. In the decision-level fusion stage, we employed maximum confidence, weighted average, and rule-based fusion strategies to comprehensively evaluate their contributions to the recognition performance. Experimental results demonstrate that the proposed approach significantly improves recognition accuracy.