Underwater target recognition based on auditory electroencephalogram features

To address the challenges in accurate underwater target recognition and leverage the advantages of hu⁃ man auditory perception in handling target recognition tasks in complex environments,an approach utilizing audi⁃ tory electroencephalogram(EEG)signals evoked by underwater acoustic targets is proposed. Experimental re⁃ sults demonstrate that the ECA-CNN network model based on a channel attention mechanism achieves a recognition rate of 94. 35% for EEG signals elicited by four types of ship radiated noise,significantly outperforming methods such as Riemannian manifolds and support vector machines. This indicates that auditory attention EEG features can effectively identify underwater acoustic targets. Moreover,by incorporating insights from neuroscience re⁃ search,the feature extraction results of the channel attention mechanism are interpreted,enriching the interpret⁃ ability of deep learning methods.