Abstract: In order to restrain the problem of end effect in the Empirical Mode Decomposition （EMD） process of the EEG data, a new method is proposed. This method combines Support Vector Machine （SVM） and window function to deal with the original signal. The method consists of three steps: the signal is forward and backward extended respectively for several extreme points by SVM; window function is added to the extended data; the empirical mode decomposition is used to process the original signal, and the extended signal by the proposed method respectively, and then remove the extended points from the intrinsic mode functions. In order to analyze the performance of the proposed method, orthogonality is chosen as the quantitative evaluation index. The artificial and real EEG signals are used as experimental objects of simulation experiments. Comparing with the other methods, the experimental resluts show that the proposed method can effectively restrain the end effects during the process of empirical mode decomposition.

Key words: empirical mode decomposition, end effects, continuation, support vector machines, window function, orthogonality

摘要： 针对经验模态分解在对脑电数据进行处理时存在的端点效应问题，提出了一种新的端点效应抑制方法。该方法将支持向量机和数据加窗进行结合对原始信号进行处理。该方法包括三个步骤：采用支持向量机对原始信号两端分别延拓有限个极大值和极小值;用窗函数对延拓后的数据进行加窗处理;分别对原始信号以及支持向量机延拓和加窗处理后的信号进行经验模态分解，并舍弃各阶固有模态函数中延拓的数据点。为了分析所提方法的性能，以正交性作为量化评价指标对比不同方法的性能。以人工信号和实际脑电信号为实验对象进行的模拟实验表明，相比于其他几种方法，提出的方法可有效抑制经验模态分解处理过程中端点效应问题。

关键词: 经验模态分解, 端点效应, 延拓, 支持向量机, 窗函数, 正交性