Underwater Acoustic Target Recognition Based on Depthwise Separable Convolution Neural Networks.

Gang Hu,Kejun Wang,Liangliang Liu

doi:10.3390/s21041429

Abstract

Facing the complex marine environment, it is extremely challenging to conduct underwater acoustic target feature extraction and recognition using ship-radiated noise. In this paper, firstly, taking the one-dimensional time-domain raw signal of the ship as the input of the model, a new deep neural network model for underwater target recognition is proposed. Depthwise separable convolution and time-dilated convolution are used for passive underwater acoustic target recognition for the first time. The proposed model realizes automatic feature extraction from the raw data of ship radiated noise and temporal attention in the process of underwater target recognition. Secondly, the measured data are used to evaluate the model, and cluster analysis and visualization analysis are performed based on the features extracted from the model. The results show that the features extracted from the model have good characteristics of intra-class aggregation and inter-class separation. Furthermore, the cross-folding model is used to verify that there is no overfitting in the model, which improves the generalization ability of the model. Finally, the model is compared with traditional underwater acoustic target recognition, and its accuracy is significantly improved by 6.8%.

Highlights

Underwater acoustic target recognition technology is used to analyze ship radiated noise received by sonar and to judge the classification of the target [1,2], which has important economic and military value
Traditional underwater acoustic target recognition methods based on ship radiated noise classify ship types by using artificially designed features and shallow classifiers, focusing on feature extraction and the development of nonlinear classifiers [3,4,5,6,7,8]
We propose a new deep convolution neural network model for feature extraction and classification of ship radiated noise, which includes a series of depthwise separable convolution, fusion layer and time-dilated convolution

Summary

Introduction

Underwater acoustic target recognition technology is used to analyze ship radiated noise received by sonar and to judge the classification of the target [1,2], which has important economic and military value. The features of artificially designed ship-radiated noise include waveform [9], spectrum [10], and wavelet [11], etc., which are dependent on expert knowledge and prior knowledge and have weak generalization ability. Shallow classifiers such as support vector machine (SVM) [12] and the shallow neural network classifier [13] have weak fitting and generalization abilities when dealing with complex and large numbers of samples. In the classification model based on auditory features, auditory filter banks designed based on perceptual evidence tend to focus only on the property of signal description rather than the purpose of classification [14,15]

Methods

Results

Conclusion