Research On Target Recognition Of Underwater Acoustic Signal Based On Multi-dimensional Feature And Deep Learning

With the development of marine resources,the development of ocean exploration technology and the proposal of ocean power strategy,the research on underwater acoustic signal target recognition is also in-depth.The task of underwater acoustic signal target recognition is based on underwater acoustic signal data,through preprocessing,feature extraction and recognition processes to recognition underwater acoustic signal target quickly and accurately.Underwater acoustic signal target recognition has important theoretical and practical significance in both military and ocean exploration fields.A method with combination of multi-dimensional fusion features and modified Deep Neural Network(MFF-MDNN)is proposed to recognize underwater acoustic target in this paper.Specifically,due to complex and changeable underwater environment,single feature is difficult to describe underwater acoustic signals.Therefore,it is necessary to propose a good feature extraction method.Gammatone Frequency Cepstral Coefficient(GFCC)and Modified Empirical Mode Decomposition(MEMD)are developed to extract multi-dimensional features in this paper.On this basis,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,in order to reduce redundant features and further improve recognition accuracy,Gaussian Mixture Model(GMM)is used to modify the structure of Deep Neural Network(DNN),which proposed a new DNN based on GMM called Modified Deep Neural Network(MDNN).Compared with other methods,MFF-MDNN has better recognition effect and adaptability.In addition,MFF-MDNN is combined with the information hiding method proposed in this paper,that is,based on the association of chaotic encryption and 3-LSB(3-ACL)information hiding method,MFF-MDNN is applied in the field of identity authentication.And four different indicators are selected to prove the proposed 3-ACL method also has better security and a certain ability to resist attacks.It is further proved that the MFF-MDNN proposed in this paper has strong practical application ability.