摘要
为得到区分左右手运动想像脑电信号的最优特征,提出了一种自适应单次脑电特征提取方法.该算法先按运动想像电位生理学原理对不同被试寻找事件相关去同步/同步(ERD/ERS)现象最明显的频段与时间段,再按照这些参数提取C3,C4导脑电信号的能量,最后取其能量比值作为左右手想像分类的特征.采用公共标准数据集做测试,运用支持向量机(SVM)进行分类,并与AR特征提取法对照.结果表明,该法可有效提高分类正确率(平均90.7%,最佳98.7%),优于使用固定频段与时间段的AR特征提取法(平均77.4%,最佳92.8%),且算法复杂度低于AR特征提取法,适应性稍强于AR特征提取法,适合在线应用.
A features extraction method is proposed to find the optimum features of hand motor imagery(MI).First,the subject-specific discriminative frequency and the time range that show the most dominant event related desynchronization/synchronization(ERD/ERS) phenomenon is selected.Then according to these information,the energy of EEG signal(C3,C4) is extracted.Finally,the ratio of the C3 energy to the C4 energy is adopted as the optimum feature.For test purpose,features are extracted from the standard date set and support vector machine(SVM) is used as the classifier.The proposed method can improved the classification accuracy(the average is 90.7%,the optimum is 98.7%),which is better than the existing autoregression model(AR) method(the average is 77.4%,the optimum is 92.8%),and it is suitable for the online analysis.
出处
《东南大学学报（自然科学版）》
EI
CAS
CSCD
北大核心
2009年第S1期219-223,共5页
Journal of Southeast University：Natural Science Edition
基金
国家自然科学基金资助项目(60513005
60674089)
上海市重点学科建设资助项目(B504)
关键词
运动想像
事件相关去同步/同步
特征提取
自回归模型
motor imagery
event related desynchronization/synchronization
feature extraction
autoregression model
