论文部分内容阅读
脑-机接口(BCI)中常用高密度导联来获取脑电(EEG)信号的空间信息,为了避免使用过多导联给EEG采集工作带来不便,消除无关的噪声通道,本文提出了一种基于共空间模式(CSP)的导联优化方法,基于CSP方法得到的投影矩阵,使用2-范数的导联筛选准则,筛选出在投影空间中权重较大的M个导联,目的是用较少的导联来获得与使用高密度导联相近的分类识别率。实验数据使用BCI Competition 2005DatasetⅢa,针对三个受试者的三类运动想象(左手、右手和脚),分别比较了使用该方法选择的导联和使用全部导联情况下得到的分类识别率。实验表明,使用筛选后的20导联得到的三个受试者的分类识别率,均高于使用全部60导联得到的分类识别率,从而验证了所提出方法的有效性和实用性。
In order to avoid the inconvenience of using EEG signal and eliminate the unrelated noise channel by using too many leads, a high-density lead is usually used to acquire the spatial information of EEG signals in the brain-computer interface (BCI) Based on CSP-based lead optimization method, based on the projection matrix obtained by CSP method, 2-norm lead screening criterion is selected to screen out M lead with larger weight in projection space. The purpose is Use fewer leads to get a classification recognition rate that is similar to using high-density leads. EXPERIMENTAL DATA BCI Competition 2005 Dataset IIIa was used to compare three types of motor imagination (left hand, right hand, and foot) for three subjects, respectively comparing the leads selected using this method with those using all leads. Experiments show that the classification accuracy of the three subjects obtained by using the filtered 20 leads is higher than that of all the 60 leads, which verifies the effectiveness and practicability of the proposed method.