Recognition of overlapped Radar signals via VMD-Based Multi-Label Learning

In complex electromagnetic environment, it is more likely to receive overlapped radar signals. In order to recognize the individual modulation types of these signals, this paper proposes a novel multi-label learning framework consisting of two modules. The first one is VMD-based feature extraction, which is effective and flexible to reconstruct original signals while maintaining useful information. The second one is complex deep network which can be trained simply with single-modulated signal samples. Experiments have demonstrated that the proposed method can achieve better performance than other methods under low signal-to-noise ratio conditions.