基于能量聚类分析法的运动想象脑电信号快速识别

在脑-机接口研究中,针对两种思维任务的特征提取和分类,提出一种以单次实验的信号能量与其中两个通道信号的平均能量的欧式距离作为特征,用聚类分析法进行任务分类的方法,7-折交叉验证法用来评价分类器的性能;采用BCI 2003竞赛数据集Ⅲ,分析了该方法的实验背景和理论依据,并将分类精度与竞赛的结果进行了比较;表明了所提出方法运用在实际系统中的有效性。

Fast Recognition of Motor Imagery EEG Based on Energy Cluster Analysis Method

In the study of brain-computer interfaces, a method of feature extraction and classification used for two kinds of imaginations was proposed. Euclidean distances between mean energies recorded from channels of the two kinds and the signal energy of the single trial were considered as features, and cluster analysis method was used to determine imagination classes, 7-fold cross validation was used to evaluate the performance of classifier. The background of experiment and theoretical foundation were analyzed referring to data sets Ⅲ of BCI 2003 competition, and the classification error was compared with the results of the competition. It is showed that the method has a better efficiency for applying to practical systems.