Feature Extraction and Classification of Ship Targets Based on Gammatone Filter Bank

Feature extraction is the key technology in underwater target recognition. The complexity of underwater environment and targets makes it very difficult to extract accurate features of underwater acoustic targets. This paper proposes a method based on the Gammatone filter bank to extract the feature vector of the radiated noise of the underwater ship target. The radiated noise signal is filtered by the Gammatone filter bank, and the dyadic discrete wavelet transform is used to remove the background noise interference and retain the local detailed information. The GDWC (Gammatone Discrete Wavelet Coefficient) eigenvectors of the signals were obtained, and the target eigenvectors were classified and discriminated by K-nearest neighbor algorithm. The experimental results show that the recognition effect of the proposed GDWC auditory feature extraction method is better than that of the MFCC (MEL Frequency Cepstral Coefficient) feature extraction method and its improved algorithm CMFCC (Compound MEL Frequency Cepstral Coefficient), and has good robustness, which can be applied to underwater target identification.