基于EMD与CSP算法的运动想象脑电信号分类

Classification of Motor Imagery Electroencephalogram Based on EMD and CSP Algorithm

针对传统时频分析方法在提取脑电信号特征时利用空间信息少、分类准确率低的问题，提出了一种基于经验模式分解（EMD）和共空间模式（CSP）融合算法的脑电信号特征提取方法。在原有EMD算法提取的时频特征的基础之上，加入经CSP滤波后的空域特征，构成一个新的特征向量，再用支持向量机（SVM）进行分类。采用该方法对第三届国际BCI竞赛datasetⅢ进行处理，分类准确率为92．85％，相对于单纯的EMD算法，准确率提高了5．7％。实验结果表明，所提出的方法能够有效地提取脑电信号的时-频-空域特征，提高分类准确率。

Considering the fact that there exist some problems in the traditional time-frequency analysis method when extracting the feature of EEG, such as the lack of spatial information and low classification accuracy, a new feature extraction method based on empirical mode decomposition （EMD） and common spatial pattern （CSP） is proposed. In this work, the spatial feature of EEG after CSP filtering is added into the time-frequency feature extracted by empirical mode decomposition, and a support vector machine （SVM） is used to classify the extracted feature. The classification accuracy is 92.85% when we adopt the proposed method to deal with the dataset Ⅲ of the third international BCI competition. Compared with pure EMD algorithm, the method proposed improves the accuracy by 5.7%. The experiment results show that the method can extract the time- frequency-space domains＇ feature of EEG in the time-frequency-space domain effectively and improve the classification accuracy.