小波分解和改进卷积神经网络相融合的水声目标识别方法

An improved convolutional neural network (CNN) combined with wavelet decomposition was developed for the classification and recognition of underwater acoustic signals with non-stationary characteristics. An underwater target recognition algorithm called WAVEDEC_CNN was developed and verified using four types of collected lake test data. Compared with the traditional MFCC+SVM method, the WAVEDEC_CNN algorithm increased the correct recognition rate by 15.38%. Additionally, compared with the NO_CNN, WPDEC _CNN and EMD _CNN methods, the correct recognition rate of the WAVEDEC_CNN algorithm was increased by 4.41%, 3.23%, 12.81%, respectively. Furthermore, the proposed WAVEDEC_CNN algorithm had the shortest calculation time compared with the other methods. These results show that the proposed method can be effectively applied in underwater acoustic target recognition.