Abstract: In order to solve the endpoint effect in the empirical mode decomposition, an improved method combining the mirror extension with the Support Vector Regression（SVR） is proposed. In the improved method, the SVR method is applied to predicting extreme points on both ends of the extreme points of the original signal, and then the mirror extension method is applied to determining the position of the predicted extreme points. The improved method can solve the inaccurate prediction on the long data sequence by using the SVR method separately, and the problem that the boundary of the short time sequence is not the extreme point by using the mirror extension method separately. At the end, the endpoint extension effect is analyzed by using the evaluation criteria. The simulation results show that the improved method can effectively restrain the endpoint effect of the empirical mode decomposition.

Key words: empirical mode decomposition, endpoint effect, mirror extension, Support Vector Regression（SVR）

摘要： 针对经验模态分解过程中产生的端点效应问题，提出了将镜像延拓和支持向量回归机相结合的端点延拓改进方法。利用支持向量回归机对原始信号的极值点数据序列两端进行预测，用镜像延拓法确定所预测极值点的位置。该改进方法解决了支持向量回归机对长数据序列预测不准确，以及镜像延拓法对端点不是极值点的短数据序列处理效果不佳等问题。引入六个评价标准，对端点延拓方法的效果进行了分析。结果表明，该改进方法能有效地抑制经验模态分解产生的端点效应。

关键词: 经验模态分解, 端点效应, 镜像延拓, 支持向量回归机