一种基于相关系数的CSP成分自动选取方法

An Automatic Component Selection Method for Common Spatial Pattern Based on Correlation Coefficient

在基于运动想象的脑机接口系统中,共同空间模式方法作为一种有效的分类处理方法已被广泛运用。使用该方法时,选择合适的成分构造滤波器是非常重要的步骤,直接影响到分类准确率。该文提出了一种基于相关系数的共同空间模式滤波器成分自动选择方法,通过使用2008年脑机接口竞赛数据检验,平均分类准确率明显高于使用传统成分选择方法构造的滤波器,验证了该方法的有效性。

In brain-computer interface （BCI） systems based on motor imagery, the common spatial pattern （CSP） has been widely used as an effective classification method. It＇s very important to choose proper component to build the spatial filter, which affects the classification accuracy directly. In this paper, it proposed a method based on the correlation coefficient about how to choose these components automatically. By testing this method with datasets from BCI Competition 2008, it got higher average classification accuracy than that of the traditional method, proved its efficiency.