A Broadband Underwater Acoustic Signal Recognition Method Based on Information Fusion

The oceans are filled with a variety of valuable underwater acoustic signals, including active sonar signals, underwater communication signals, marine mammal signals, and ship-radiated noises. In this paper we investigate the recognition method for these four categories of signals. First of all, we design an one-stage convolutional detection network based on spatial attention mechanism,. Furthermore, an information fusion strategy is proposed to address the limitations of LOFAR spectrum in representing low-frequency signals and Mel spectrum in representing high-frequency signals. This approach significantly enhances the overall recognition efficacy for the four categories of signals.