Feature Extraction Technique of Acoustic Target Based on Wavelet Packet Energy and Principal Component Analysis

摘要

A feature extraction method based on wavelet packet energy distribution and correlation coefficient has been put forward to recognize the different acoustic targets in this paper. In view of the characteristics of acoustic target, we employed principal component analysis (PCA) to compress data set of the features extracted based on wavelet packet energy distribution and correlation coefficient. The results have been inputted into the neural network as eigenvectors for pattern recognition. Simulation results indicate that the method suggested in this paper have a recognition rate better than 8% with only wavelet packet energy method, thus verifying its effectiveness.