Feature-extraction And Target Recognition On Underwater Ship-radiated Acoustic Noise

Since the 21st century,the status of the ocean in a country,especially its military strategic status,has become increasingly prominent.During the development and utilization of marine resources,in order to maintain safety in this process,whether it is civilian or military,the underwater acoustic targets must be correctly identified and tracked.In order to better monitor a particular sea area,it is also necessary to classify and identify underwater acoustic targets.Therefore,based on the recognition of underwater acoustic technology,this paper takes the shipradiated noise as the main research object to study the recognition of different types of ships.Although underwater acoustic target recognition technology has been developed for decades,in many cases it still cannot satisfy the actual demands.Therefore,based on previous research,this paper explores and studies novel ship-radiated noise recognition technology in combination with current advanced pattern recognition knowledge to further improve ship-radiated noise recognition performance.The research contents of this paper mainly include three aspects.The first part is to analyze the generation mechanism of ship-radiated noise,study the spectral characteristics and underwater acoustic characteristics of ship-radiated noise,and finally describe the recognition method of ship-radiated noise.The second part of the research is to explore the technology of ship-radiated noise recognition based on a pattern recognition method(feature fusion).This part of the research first extracts multi-dimensional features suitable for underwater acoustic target recognition,including temporal features,spectral features and cepstrum features,and then introduces principal component analysis and canonical correlation analysis techniques to propose a novel ship-radiated noise recognition method based on feature fusion.Then we apply the proposed method to real ship-radiated noise data to verify the effectiveness of the proposed method.The third part of the research is to explore the unsupervised ship-radiated noise recognition technology.This part of the research mainly explores deep learning method for ship-radiated noise recognition,introduces unsupervised learning technology in pattern recognition,proposes an unsupervised ship-radiated noise recognition method,and finally validates the method on a real ship-radiated noise database.The experiment results indicates that the proposed method can improve the accuracy of ship-radiated noise recognition with great effectiveness.