Underwater Acoustic Target Recognition Based on Dual Attention Networks and Multiresolution Convolutional Neural Networks
Chengwei Liu, Feng Hong, Haihong Feng, Menglu Hu
Abstract
Underwater target acoustic recognition (UATR) based on radiated noise is one of the main applications of the passive sonar system. However, the factors such as the complexity of the marine environment, the diversity of target types, and the scarcity of data make it difficult to get a high classification accuracy in most of the UATR systems. To solve this problem, a novel method based on Dual Attention Networks (DAN) and a Multiresolution Convolutional Neural Network (DAN-MCNN) is proposed to solve this problem. Firstly, we designed three-dimensional (3D) aggregated features based on the multi-class feature subsets, which are composed of MFCC, Log-mel Spectrogram, Chroma, Spectral Contrast, and Tonnetz. Then, based on the frequency perception mechanism of the human ear and auditory attention mechanism. A multi-resolution pooling and convolution scheme is adopted to construct the MCNN architecture, which can better adapt to the time-frequency structure of the 3D aggregated characteristics. Besides, we propose the DAN module consist of Position Attention Module and Channel Attention Module. The dynamic weighting method is used to emphasize the regions of interest and suppress the irrelevant background regions, to capture the global dependence and local characteristics of samples. The experimental results show that the proposed approach achieved average recognition accuracy of 95.6% in the ShipsEar dataset, which can effectively improve the recognition accuracy of underwater targets and get the best classification result.