Recognition of oceanic explosion signals based on wavelet packet transform and neural network

A recognition method of hydroacoustic signatures of explosions conducted underwater and in the low atmosphere based on wavelet packet transform and neural network is presented. Characteristics of hydroacoustic signature were analyzed and the mainly multi-scale energy feature vector was extracted by wavelet packet decomposition method. Then the neural network was trained and optimized by the imported feature vector and the trained network acted as the classifier to recognize and classify the hydroacoustic explosion signal. The experimental result shows the efficiency and credibility of this method.