A Method of Underwater Acoustic Signal Classification Based on Deep Neural Network

As an important precondition for underwater acoustic signal analysis and underwater target detection & identification, classification of underwater acoustic sensor signal (especially in low signal-to-noise ratio) has become a research focus on underwater acoustic sensor signal processing in terms of complicated acoustic environment and increasingly slighter target noise. Mel-Frequency Cepstrum Coefficients (MFCC) of underwater acoustic signal are extracted in the Paper to classify and identify characteristics spectrogram of MFCC of underwater acoustic signal in combination with recurrent neural network and convolutional neural network and a Deep Neural Network (DNN) model is established for classification of underwater acoustic signal to propose a method of underwater acoustic signal classification based on DNN, support efficient classification of underwater acoustic signal, and provide high-quality data input for underwater acoustic signal analysis and underwater target identification.