解相关EMD:消除模态混叠的新方法  

Decorrelation EMD: a new method of eliminating mode mixing

作　　者：肖瑛[1,2] 殷福亮[1] 

机构地区：[1]大连理工大学电子信息与电气工程学部,辽宁大连116024 [2]大连民族学院信息与通信工程学院,辽宁大连116600

出　　处：《振动与冲击》2015年第4期25-29,共5页Journal of Vibration and Shock

基　　金：国家自然科学基金(61201418);中央高校基本科研业务费专项资金资助(DC12010218);辽宁省高等学校优秀人才支持计划(LJQ2013126)

摘　　要：模态混叠是制约希尔伯特-黄变换(Hilbert-Huang Transform,简称HHT)在工程上推广应用的一个主要难题,采用在经验模态分解(Empirical Mode Decomposition,简称EMD)过程中嵌入解相关操作有效解决了EMD过程中产生的模态混叠问题。模态混叠的本质是各本征模态函数(Intrinsic Mode Function,简称IMF)分量之间信息相互耦合,无法表征有意义的物理过程,因此可认为模态混叠现象是由于IMF之间不完全正交导致。根据零均值随机信号正交性与不相关具有等价关系,在EMD过程中嵌入解相关操作,可有效保证IMF之间的正交性,消除模态混叠。仿真信号处理结果证明了方法的有效性。Mode mixing is one of difficulties of HHT in engineering applications. Here, a new method embedding a decorrelation operator into EMD process for solving mode mixing was proposed. The essence of mode mixing was mutual coupling of IMFs, it led to that a meaningful physical process could not be expressed with these IMFs, and it was considered that the mode mixing was due to that IMFs were not completely orthogonal. As the orthogonality and the non- correlation of a random process were equivalent, embedding a decorrelation operator into EMD process could improve the orthogonality of IMFs effectively and eliminate mode mixing. The simulation results showed the effectiveness of this method.

关 键 词：经验模态分解 解相关 正交性 时频分析 

分 类 号：TN957.52[电子电信—信号与信息处理]