Adaptive wavelets classification of transient sonar signals

This paper discusses the applicability of adaptive wavelets for the classification of transient sonar signals. A two-step classification method is presented. The first step is the extraction of adaptive wavelet features. The second step is the signal classification using a feedforward neural network. Four classes of transient sonar signals are used for an experiment. The test result shows that the performance of adaptive wavelets for this application is rather better than that of the power spectral features based classifier.