Ship Radiated Noise Classification Method Based On Deep Neural Network

The classification and recognition technology of underwater acoustic targets has always been one of the most important research contents in the field of underwater acoustic signal processing.The purpose of feature extraction is to study and select stable eigenvectors that can effectively reflect the target signal class.The goal of classifier design is to design a classifier that can accurately segment different target eigenvectors according to the characteristics of different classifier models.The research object of this paper is ship radiation noise,and the research focuses on feature extraction algorithm and classifier design.The feature extraction technology and deep neural network classification method used in the speech signal processing algorithm are introduced into the classification and identification research of ship radiation noise.The main contents of the paper are as follows:Firstly,this paper introduces the generation mechanism of ship radiation noise and common noise types.The ship radiation noise is simulated,including noise broadband continuum,narrow-band line spectrum,beat modulation function,marine environmental noise,etc.Three different kinds of ship radiation noise signals is simulated altogether.Secondly,the feature extraction of the ship's radiation noise signal is carried out.Analyze the mathematical principles of DEMON spectral analysis,1.5-dimensional spectral feature extraction,and Mel-Cepstral Coefficient(MFCC)feature extraction method,and simulate the three feature extraction methods.Then SVM and CNN are used as classifiers to classify ship radiation noise signals.The mathematical principle of SVM and the basic structure and working principle of CNN are analyzed theoretically.The support vector machine model and convolutional neural network model are built by using MATLAB platform to classify the target features.By comparing the classification effects of the two classifiers under different SNRs,the performance of the classifier and the advantages and disadvantages of the feature extraction methods are analyzed.