基于二阶盲辨识的运动想象脑电信号分类

Classification of motor imagery electroencephalogram signals based on second-order blind identification

人类的思维活动会引起脑电信号的变化,而对特定的脑电信号的分析可以设计脑机接口通信系统。文章提出了一种应用于脑机接口系统的运动想象脑电信号分类方法,即采用一种盲源分离算法——二阶盲辨识对运动想象脑电信号进行处理,进而根据不同类型间的Fisher距离提取特征,最后运用多层BP神经网络进行分类,3个样本的平均分类准确率达到90.9%,而直接对原始脑电信号进行分类,其平均分类准确率仅为72.6%,这表明二阶盲辨识算法的运用明显改善了运动想象脑电信号的分类准确率。不同受试者及不同训练样本的分类准确率表明,以BP神经网络作为分类器,其结果与训练样本关系密切,尤其是在样本数较少的情况下。从实时性方面来说,二阶盲辨识算法完全满足要求。

Human thinking activities evoke electroencephalogram （EEG） signal changes, so EEG analysis can help to design communication systems of brain-computer interface （BCI）. In this study, we developed a motor imagery classification strategy for BCI applications. Second-order blind identification （SOBI）, a blind source separation （BSS） algorithm, was applied to preprocess EEG data. Subsequently, Fisher distance was used to extract certain features. Finally, classification of motor imagery EEG signals was performed by back-propagation （BP） neural networks. The average classification accuracy was 90.9% with SOBI preprocessing but 72.6% without SOBI preprocessing, indicating that classification accuracy of motor imagery EEG signals was significantly improved by SOBI preprocessing. Classification accuracy of different subjects and different training sets shows the result is closely related with training set when BP neural network is used as classifier, especially the number of samples is less. From real-time context, SOBI algorithm fully meets the requirements.