基于复合肢体动作观察的下肢运动想象脑电识别  

Lower limb motion image electroencephalogram recognition based on compound limb motion observation

作　　者：姜晓宏 董洪涛[1] 张广举[1,2] 孙国栋 孟琳 JIANG Xiaohong;DONG Hongtao;ZHANG Guangju;SUN Guodong;MENG Lin(Academy of Medical Engineering and Translational Medicine,School of Medicine,Tianjin University,Tianjin 300072,China;Haihe Laboratory of Brain-computer Interaction and Human-machine Integration,Tianjin 300000;The Third Affiliated Hospital of Shandong First Medical University,Affiliated Hospital of Shandong Academy of Medical Sciences,Jinan 250031,China)

机构地区：[1]天津大学医学院医学工程与转化医学研究院,天津300072 [2]天津市脑机交互与人机共融海河实验室,天津300000 [3]山东第一医科大学第三附属医院,济南250031

出　　处：《生物医学工程研究》2024年第6期432-438,共7页Journal Of Biomedical Engineering Research

基　　金：国家重点研发计划项目(2022YFF1202500,2022YFF1202503)。

摘　　要：针对脑电(electroencephalogram,EEG)信号的下肢运动想象(motor imagery,MI)脑机接口系统受限于下肢MI信号弱隐、MI诱发范式不够自然或有效性不佳等问题,本研究设计了一种站立姿态下的复合肢体MI任务。首先,本研究在站立姿态下进行了复合肢体的MI实验,并使用视频辅助降低MI难度,提升MI效果;其次,利用共空间、滤波器组共空间及特定受试者的共空间模式提取EEG特征,并使用基于互信息的个体最佳特征选择算法进行特征选择;最后,利用支持向量机(support vector machine,SVM)进行分类。本研究采集了12名受试者的EEG数据并进行了两侧复合肢体MI EEG的二分类,以及两侧单下肢MI EEG的二分类,利用特定受试者的共空间模式进行特征提取时,所得复合肢体MI的平均分类准确率为0.70±0.06,较单下肢MI高6%。该结果验证了复合肢体MI范式的EEG诱发效果较单侧肢有一定的提升作用。Aiming at the problems that the lower limb motor imagery(MI)brain-computer interface system based on electroencephalogram(EEG)signals was limited by weak and implicit lower limb MI signals,and the MI evoked paradigm was not natural or effective,we designed a composite limb MI task in standing posture.Firstly,the limb MI experiments of composite limbs was conducted in the standing posture,and the video assistance was used to reduce the difficulty of MI and enhance the effectiveness of MI.Secondly,the EEG features were extracted by common spatial patterns(CSP),filter bank common spatial patterns(FBCSP),and subject-specific common spatial patterns(SSCSP),and the individual best feature selection algorithm based on mutual information was used for feature selection.Finally,the support vector machine(SVM)was used for classification.On the basis of the EEG data collected from 12 subjects,binary classification of compound limb motor imagery and binary classification of unilateral lower limb MI were conducted.When using SSCSP for feature extraction,the average classification accuracy for compound limb MI was 0.70±0.06,which was 6%higher than that for unilateral lower limb MI.It proves that the EEG induce effect of the composite limb movement imagination paradigm can improve the unilateral limb.

关 键 词：运动想象 脑机接口 复合肢体 运动意图识别 

分 类 号：R318[医药卫生—生物医学工程] TP183[医药卫生—基础医学]