Acoustic target identification technology copes with the acoustic singals produced by moving targets in the medium,and extract the feature vectors,then finally recognize the targets. On account of passively receiving acoustic signals and detecting, passive acoustic identification has good concealment and is comprehensively applied in modern military wars.This paper based on multi-scale resolution, discuss the three battlefield target signals by denoising the signals extracting feature vectors and designing of classifiers in order to targets recognition. Acoustic target identification technology has three steps:above all, denoising the signals, secondly feature vectors extraction,thirdly designing classifier. As a result, the paper does some research in denoising methods of different threshold rules and analyzes the effect of different threshold rules by comparative experiments. At last, we get some experience of denosing battlefield acoustic signals. Then, we take advantage of MFCC and wavelet packet analysis in extracting feature vectors. Finally, we apply genetic algorithm in BP neural network and optimize the threshold values in order to avoid the defect of part least values.at last, it proves that the method is feasible and improves targets recognition rate largely.In a word, the methods the paper provides have highly targets recognition rate and give some reference in reearch of acoustic target identification technology in the future. |