经验模态分解的端点问题数值处理方法研究

Numerical processing method of endpoint issue of EMD

经验模态分解(EMD)是由Huang等发展的一种新的数据分析方法,但在利用样条插值获得上下包络过程中存在着棘手的端点问题。文章在该问题已有解决方法的基础上,提出了基于极值点单调性一致的EMD端点问题处理方法。根据信号的极值序列查找与数据末端极值的差值和同时具备最小、单调性一致且在单调性内的点数相等三个条件的极值序列,进而构造方程组进行极值预测。通过与其他两种方法的对比验证,证明了提出的方法可以有效抑制端点效应。

The Empirical Mode Decomposition （EMD） is a new method for data analysis developed by Huang, etc. But there is a troublesome endpoint issue during the course of obtaining two envelops of the data with spline interpolation. A new endpoint issue processing method based on same monotonicity of extreme points is proposed on the basis of existing algorithms. The difference values between the extremums of data end are looked up according to the extremum sequence which satisfies the conditions simultaneously such as minimum value, same monotonicity and equal point count in the monotonicity zone. Then the extremums are predicted with constructed equation set. It is verified that the proposed method can restrain endpoint effect availably by the comparison experiments with the other two methods.