基于共空间模式的运动想象脑电信号识别研究

摘要

Classification of electroencephalogram(EEG)signal is an important issue in brain-computer interface(BCI). Based on the classification of the EEG signals,in this paper,we collect the left-right hand motor imagery EEG data of 7 subjects which are recorded by EGI-64 scalp electrodes placed according to the international 10/20 system. Firstly,the EEG data are denoised with extend Infomax-Independent Component Analysis ( ICA );Secondly, C3 and C4 electrodes features are extracted by using Common Spatial Pattern( CSP);Finally,the average classification rates of Fisher Linear Discriminant Analysis(FLDA),Bayesian,Radial Basis Function(RBF)neural network and BP neural network methods are compared. The classification results show that the average classification rate of neural network is higher than the other two methods,and that the average classification rate of BP neural network can be up to 95. 36%,but the other three methods of running velocity is obviously faster than the BP neural network. The results provide a basis for real-time BCI system implementation.%脑-机接口技术领域的关键问题是脑电信号的分类识别研究.本文针对脑电信号的分类问题，基于EGI-64导脑电采集系统得到7名被试者的左右手运动想象脑电数据，首先采用扩展Infomax-ICA方法对脑电数据进行去噪处理；然后利用共空间模式方法对C3/C42个电极的脑电信号进行特征提取；最后比较了Fisher线性判别分析法、贝叶斯方法、径向神经网络和BP神经网络几种算法的平均分类率.结果表明：神经网络分类方法得到的平均分类率要高于其他2种方法，而BP神经网络方法的平均分类率最高，可以达到95.36%，但另外3种方法的运行速度明显高于BP神经网络.该结果为实时BCI系统实施提供了一定依据.