基于汉字默读的运动想象脑电信号识别研究  

Research on MI-EEG recognition based on Chinese characters silent reading

作　　者：金铭 郭苗苗[1,2] 李梦凡 蔡梓良 JIN Ming;GUO Miaomiao;LI Mengfan;CAI Ziliang(State Key Laboratory of Reliability and Intelligence of Electrical Equipment,School of Electrical Engineering,Hebei University of Technology,Tianjin 300130,China;Tianjin Key Laboratory of Bioelectromagnetic Technology and Intelligent Health,School of Electrical Engineering,Hebei University of Technology,Tianjin 300130,China)

机构地区：[1]河北工业大学电气工程学院省部共建电工装备可靠性与智能化国家重点实验室,天津300130 [2]河北工业大学电气工程学院天津市生物电工与智能健康重点实验室,天津300130

出　　处：《现代电子技术》2022年第10期137-141,共5页Modern Electronics Technique

基　　金：国家自然科学基金项目(51707054);国家自然科学基金项目(61806070);河北省自然科学青年基金项目(F2018202088)。

摘　　要：基于运动想象的脑机接口(BCI)系统被广泛用于医疗康复领域,但随着想象任务的增多,分类准确率会急剧下降。为提高个体多类运动想象脑电信号(EEG)的识别准确率,文中设计附加汉字默读(4个汉字)的运动想象实验范式,采用共空间模式(CSP)+Fisher分类器和卷积神经网络两种方式对脑电信号进行特征提取并分类,验证附加默读对分类准确率的影响。结果表明:CSP+Fisher分类器方式下,附加默读的运动想象任务平均分类准确率(37.78%±3.46%)优于仅进行运动想象任务时的平均分类准确率(39.73%±3.15%),P>0.05;卷积神经网络(CNN)方式下附加默读的运动想象任务分类准确率(59.13%±2.95%)显著优于仅进行运动想象任务时的分类准确率(62.60%±2.41%),P<0.01。因此,文中提出的基于汉字默读的运动想象实验范式可以在受试者想象运动时获得更加稳定的脑电信号,从而提升BCI的分类准确率,为BCI范式的优化提供依据。The brain-computer interface(BCI)system based on motion imagery(MI)is widely used in the field of medical rehabilitation. However, with the increase of the number of imagination tasks, the classification accuracy would decline dramatically. In order to improve the recognition accuracy of electroencephalography(EEG)signals when subjects perform multiclass motor imagery tasks,the MI experimental paradigm adding Chinese characters silent reading(4 Chinese characters)is designed,the common spatial pattern(CSP)+ Fisher classifier and the convolution neural network(CNN)are used to extract and classify EEG feature,to verify the effect of additional silent reading on classification accuracy. The results show that:in the way of CSP+Fisher classifier,the average classification accuracy of MI with additional silent reading is better than that of MI task only(37.78%±3.46% vs 39.73%±3.15%,P>0.05). In the CNN mode,the classification accuracy of motor imagination task with additional silent reading is significantly better than that of motor imagery task only(59.13%±2.95% vs 62.60%±2.41%,P<0.01). Therefore,the MI experimental paradigm based on Chinese character silent reading proposed in this paper can obtain more stable EEG when subjects imagine motion,thus improving the classification accuracy of BCI,and providing the basis for the optimization of BCI paradigm.

关 键 词：运动想象 脑电信号 汉字默读 卷积神经网络 共空间模式+Fisher分类器 脑机接口 

分 类 号：TN911.1-34[电子电信—通信与信息系统] R318[电子电信—信息与通信工程]