Underwater Acoustic Target Recognition based on Graph Wavelet Bispectrum

It is challenging to recognize underwater acoustic targets because of the complex marine environment. An analysis approach based on the graph wavelet bispectrum (GWB) is proposed to handle the problem of challenging feature extraction. After the pre-processing step, the graph signals are constructed based on the area of the vertex triangle. The GWB is performed with the components of the graph signal at different scales. Five spectral features from each GWB are extracted to form the feature vector. The obtained feature vectors are categorized by SVM. The experimental results have demonstrated that the proposed method, based on the 5-fold cross-validation, achieves a high classification accuracy of 98.56%.