摘要
本研究提出一种从单次试验的多导EEG信号中提取运动相关去同步化和同步化电位特征的空间模型 ,区分左右手想象运动 ,作为一种新的通讯手段对外界设备进行控制。此模型根据各电极对分类的重要性自动获得其权值 ,并将EEG信号沿最适合分类的几个方向投影 ,沿投影方向计算一连续时间段内的方差 ,作为线性分类器的特征输入。对 8名被试者左右手想象运动时 5 9导EEG进行分类 ,正确率均在 70 %以上 ,与用多通道AR模型提取特征、神经网络做分类器的方法相比 ,效果好、速度快。
The paper proposed spatial patterns with which to extract event-related desynchronization/synchronization (ERD/ERS) potential form single trial multi-channel EEG and discriminate imagination of left and right hand movements. The spatial patterns were expected to be used as a new communication method for the disabled to control outside devices. It obtained an automatic weighting of electrodes according to their importance for the classification task. It projected EEGs onto the most discriminatory characteristic patterns and calculated the variances of a consecutive time period resulting from each projection, which were used as input features to a linear classifier. The classification correctness for eight subjects were more than 70%, which was better and faster than that with multi-autoregressive (AR) model and artifical neural networks (ANNs).
出处
《中国生物医学工程学报》
EI
CAS
CSCD
北大核心
2005年第1期85-88,共4页
Chinese Journal of Biomedical Engineering
基金
国家自然科学基金资助项目 (60 3 75 0 17)。
关键词
事件相关去同步化和同步化
空间模型
空间协方差阵
脑电分类
Automation
Biological organs
Electrodes
Handicapped persons
Neural networks
Pattern recognition
Synchronization
