基于经验模态分解和SVM的脑电信号分类方法

脑电信号的非线性、非平稳性造成对运动想象脑电信号的分类识别存在特征提取困难、可区分性低以及分类识别性能差等问题。本文提出一种基于经验模态分解(Empirical Mode Decomposition, EMD)和支撑向量机(Support Vector Machine, SVM)的运动想象脑电信号分类方法,充分利用EMD算法在处理非线性、非平稳信号的自适应性以及SVM在小样本条件的高识别性能和强泛化能力。首先利用EMD算法将C3、C4导联信号分解为一系列本征模函数(Intrinsic Mode Function, IMF),然后从IMF的信息和能量等维度提取特征将脑电信号转换至区分性更强的特征域,最后利用SVM进行分类识别。采用国际BCI竞赛2003中的Graz数据进行验证,所提方法可以得到94.6%的正确识别率,为在线脑-机接口系统的研究提供了新的思路。

Classification Method of EEG Based on Empirical Mode Decomposition and SVM

The non-linearity and non-stationary characteristics of EEG signal make it difficult to extract features and classify them into different categorizes. This paper proposes a classification method of EEG signals based on empirical mode decomposition(EMD) and support vector machine(SVM), which makes full use of the adaptive ability of EMD algorithm in dealing with nonlinear and non-stationary signals, as well as the high identification performance and strong generalization ability of SVM in small sample conditions. Firstly, C3 and C4 lead signals are decomposed into a series of Intrinsic Mode Function (IMF) by EMD algorithm, and then features are extracted from IMF"s information and energy dimensions to transform EEG signals into more discriminable feature domain, and finally SVM is utilized to classification EEG signals into different categorizes. Using Graz data from international BCI competition 2003, the proposed method can achieve a correct recognition rate of 94.6%, which providing a new idea for the study of online brain-computer interface system.