谱熵和主成分分析用于EMD分解研究  

Research on empirical mode decomposition based on spectrum entropy methods and principal component analysis

作　　者：李雪耀[1] 邹晓杰[1] 张汝波 钱真[1] 

机构地区：[1]哈尔滨工程大学计算机科学与技术学院,黑龙江哈尔滨150001

出　　处：《哈尔滨工程大学学报》2009年第7期797-803,共7页Journal of Harbin Engineering University

基　　金：国家自然科学基金资助项目(60475016);哈尔滨工程大学基础研究基金资助项目(HEUF040902)

摘　　要：针对传统经验模态分解(EMD)的终止条件的不足,将信息熵和主成分分析(PCA)原理引入希尔伯特-黄变换(HHT)中,提出了一种H ilbert时频谱熵的算法,以此为基础提出了基于时频谱熵的分量终止准则和基于主成分分析的分解终止准则,并对EMD算法进行改进.改进的EMD算法能够得到更准确的分解结果,并且在一定程度上减少了虚假分量和模态混叠的产生.仿真实验表明,采用基于时频谱熵和主成分分析的终止准则对EMD算法的改进是有效的.Information entropy methods and principal component analysis （PCA） were introduced to the Hilbert- Huang transform（HHT） method to improve performance. To resolve the disadvantages of stopping criteria in the traditional empirical mode decomposition algorithm, an algorithm using Hilbert time-frequency spectrum entropy （TFSE） was proposed. Based on this, a new stopping criterion for sifting and decomposing was presented, and the algorithm for empirical mode decomposition （EMD） was improved as well. The improved algorithm for EMD gets more accurate results and to a certain extent overcomes the problems of false mode and modal mixtures in a HHT. Simulation results showed the effectiveness of this improved method using stopping criteria based TFSE and PCA.

关 键 词：希尔伯特-黄变换 经验模态分解 筛选终止准则 主成分分析 

分 类 号：TN911.7[电子电信—通信与信息系统]