Underwater Acoustic Target Recognition: A Combination of Multi-Dimensional Fusion Features and Modified Deep Neural Network

Xingmei Wang,Yu Zhang,Fuzhao Xue,Anhua Liu

doi:10.3390/rs11161888

Abstract

A method with a combination of multi-dimensional fusion features and a modified deep neural network (MFF-MDNN) is proposed to recognize underwater acoustic targets in this paper. Specifically, due to the complex and changeable underwater environment, it is difficult to describe underwater acoustic signals with a single feature. The Gammatone frequency cepstral coefficient (GFCC) and modified empirical mode decomposition (MEMD) are developed to extract multi-dimensional features in this paper. Moreover, to ensure the same time dimension, a dimension reduction method is proposed to obtain multi-dimensional fusion features in the original underwater acoustic signals. Then, to reduce redundant features and further improve recognition accuracy, the Gaussian mixture model (GMM) is used to modify the structure of a deep neural network (DNN). Finally, the proposed underwater acoustic target recognition method can obtain an accuracy of 94.3% under a maximum of 800 iterations when the dataset has underwater background noise with weak targets. Compared with other methods, the recognition results demonstrate that the proposed method has higher accuracy and strong adaptability.

Highlights

With the development of sonar technology, underwater acoustic target recognition has become one of the major functions of sonar systems
modified empirical mode decomposition (MEMD) has been mapped to Hilbert space through Hilbert–Huang transform (HHT), the instantaneous energy (IE) and instantaneous frequency (IF) of the intrinsic mode function (IMF) were extracted, and good results have been achieved through the combination with the Mel frequency cepstral coefficient (MFCC) [16,21,22,23,24,28,29,30,31]
This paper proposes an MFF-MDNN method which combines multi-dimensional fusion features with a modified deep neural network (DNN) to recognize underwater acoustic targets

Summary

Introduction

With the development of sonar technology, underwater acoustic target recognition has become one of the major functions of sonar systems. The recognition results have been modified, it cannot accurately describe original underwater acoustic signals by only extracting one type of feature. In 2018, Sharma et al proposed a new method of combining the MFCC and modified empirical mode decomposition (MEMD) to describe mixed multi-dimensional features of original acoustic signals [16]. In 2011, Kotari et al proposed a method to recognize underwater mines, which is based on the GMM and achieved a good result [32]. This paper proposes an MFF-MDNN method which combines multi-dimensional fusion features with a modified DNN to recognize underwater acoustic targets. A dimension reduction method is proposed to ensure the same time dimension It can obtain multi-dimensional fusion features in the original underwater acoustic signals.

MMuullttii--DDiimmeennssiioonnaallFFuussiioonn FFeeaattuurreess MMeetthhoodd

Multi-Dimensional Fusion Features Algorithm

Multi-Dimensional Feature Extraction

Modified Deep Neural Network

Experiment Results and Analysis

Conclusions