基于小波包与PCA方法对水下目标识别研究

摘要

水下目标信号的分类识别一直是信号处理工程领域的研究难点.针对水下信号发声机理十分复杂与成分多样,导致表征其特征的数据量较大且维数较高,目标识别率低.要解决上述问题,需要很大的计算成本,并影响识别特性量的效率,提出了一种采用小波包与主分量分析(Principal Component Analysis,PCA)相结合的特征提取方法.通过小波包分解与重构得到水下目标辐射噪声的初始特征；用PCA方法实现对高维特征向量的优化处理.采用BP神经网络作为分类器对三类目标进行识别仿真.结果表明,减少计算量的同时,水下目标信号得到了较好的优化提取.%Underwater target classification and recognition is a research challenge of signal processing applica-tion, and as a result of complicated target signal and diverse ingredient, the character data are great and with high di-mension, which needs huge calculating cost. Under this situation, a new approach to extracting noise radiated from underwater target based on wavelet packet and principal component analysis is presented. Firstly Initial characteris-tics are obtained from underwater target by using decomposition and reconstruction of wavelet packet. Then principal component analysis is used to get the final characteristics. The final characteristics are used by designed neural net-work to recognize the noise radiated from underwater target. Experiment results show that the method of extracting features has better classification effect with low calculating cost.