二阶盲辨识算法在多类运动想象脑电信号特征提取中的应用

Research on Feature Extraction in Multi-Class Motor Imagery EEG Based on the SOBI in the Independent Experiments

为提高多任务运动想象脑电信号特征提取的准确率，在分析比较共空间模式算法的基础上，采用一种基于二阶盲辨识（Second 0rder Blind Identification，SOBI）的脑电信号特征提取方法，对四任务运动想象脑电信号进行处理，提取了能够反映不同任务类别的特征。在参考国际竞赛数据采集实验方案的基础上进行自主实验设计，采集了多名受试者的四任务运动想象脑电数据，处理结果表明：分类准确率达到了85．4％，验证了自主实验的可行性，为实现在线BCI系统奠定了实验基础，也为多任务运动想象BCI系统实现提供算法支撑。

In order to improve the accuracy of feature extraction in Multi-Class Motor Image- ry EEG in the analysis and comparison of common spatial patterns CSP algorithm. Present a EEG feature extraction method is presented based on the SOBI （Second Order Blind Identifi- cation）. Four motor imagery task EEG signals are processed to extract the different catego- ries of tasks which can reflect the characteristics. Based on the existing laboratory condi- tions,the independent experience in reference to the international competition experimental data acquisition scheme is designed. More than four subjects Imaginary motion task data are collected. The results show that the classification accuracy rate reaches 85.4% and the feasi- bility of independent experiments is verified. The support algorithm implementation for Multi-class Motor Imagery BCI system is also proposed.