Intelligent Recognition of Underwater Acoustic Target Noise by Multi-Feature Fusion

Abstract

The traditional methods of underwater acoustic target recognition need to manually extract the feature data, and the recognition process has strong human-computer interaction characteristics, which couldn't meet the demand of intelligence of underwater acoustic target recognition in the future. In this paper, a target recognition method based on Long Short-term memory network (LSTM) is proposed, which fuses multiple feature data and has a certain degree if intelligence ability. In this method, the Long Short-term memory network (LSTM) is adopted to generate various kinds of underwater acoustic target noise feature data. The SOFTMAX classifier is used to fuse multi-feature and to classify the targets. At last, this paper uses the time domain data, frequency data and Mel cepstrum data as input data to verify the model, the recognition results show that the model which fuse three kinds of input data above is better than the other methods which only one or two kinds of feature data, which meets the intelligent requirements of underwater acoustic target recognition to a certain extent.