基于CSSD和SVM脑电分类技术的研究

本文研究的BCI实验是基于BC12003竞赛数据来对脑电信号分类。本文提出了一种脑电信号趋势的概念，使用支持向量机（SVM）作为分类器的算法。首先将BC12003竞赛数据通过中值滤波器和由小波函数构成的带通滤波器，然后用时间窗进行时域上地过滤，选取对于大脑思维活动现象表现最明显的一段数据，再通过共空域子空间分解（CSSD）从脑电信号中提取特征，最后基于提取的特征，通过SVM训练后，进行分类识别，分类识别率达到了85％～96％。实验中采用的特征提取方法和分类方法对于脑电信号的分类识别准确率提高了不少。

The Study on the Classification of EEG Based on the CSSD and SVM

The BCI experiment studied in this paper used the BCI2003 competition data to classify the EEG. In this paper, we proposed a concept of EEG trend, and used support vector machine （SVM） as classification algorithms. Firstly, the multi-channel EEG data pass the low-pass filter and the band pass filter, respectively. Secondly, use the time window to filter them from time domain, select the section of the most significant phenomenon of performance data, and extract features through the common spatial subspace decomposition （CSSD） from the signals.Finally, based on the extracted features, we classify them through the SVM training. The recognition rate is 85% - 96%, and the accuracy for the EEG classification increased a lot.