基于高阶微分的EMD均值计算方法

Local Mean Computation for Empirical Mode Decomposition Based on Higher Order Derivative

为了改善经验模态分解的分离性能,提出一种基于信号高阶微分的分解算法.本文首先讨论了经验模态实现模态分离的必要条件,并证明对输入信号进行偶数阶数值微分可以提高模态分离性能.然后在此基础上提出一种以偶数阶微分的过零点为特征的均值计算方法.最后对仿真信号的分解进行了实验研究.结果表明,本文方法可以改善分离性能,性能提高的程度与理论分析结果符合;与经验模态分解相比,本文方法具有更高的分解精度.

In order to improve the separation performance of empirical mode decomposition,an algorithm based on higher order derivative is put forward. First,a necessary condition for separability is studied. It is proved that separability can be improved by even order derivative of the input signal. An algorithm that builds local mean from zero-crossings of the even order derivative is then proposed. Finally,decomposition performance of the proposed method is demonstrated by application to synthetic signals.Numerical experiments showthat the proposed method has enhanced separation quality which agrees well with the theoretical analysis. In compared with the empirical mode decomposition,the proposed method exhibits higher accuracy.