Underwater sound target identification method based on feature fusion and residual CNN (Convolutional Neural Network)

The invention discloses an underwater acoustic target recognition method based on feature fusion and a residual CNN, and the method comprises the steps: firstly extracting MFCC and GFCC features, achieving the feature dimension reduction through a domain component analysis method, fusing the two features after the dimension reduction, and inputting a residual CNN model for training. According to the method, on the basis of a ResNet18 model, a residual CNN model is built and trained, SVM, GRNN, 3CNN and ResNet18 models which independently input MFCC and GFCC features are compared, the result shows that feature fusion parameters serve as input features to train the residual CNN model, the rich information advantage of the feature fusion technology and the feature mining capacity of the residual CNN are fully played, and therefore compared with other models, the method has the advantages that the feature fusion technology is optimized, and the feature fusion efficiency is improved. Compared with the prior art, t