" Common spatial patterns analysis and linear discriminant analysis are popular algorithms for spatial filtering and classifying in motor imagery. These algorithms are very sensitive to noise and artifacts which affect the classification accuracy. To deal with this issue, it is proposed to replace the
usual estimators of covariance and scale used in the algorithms for robust versions. The performance of the methods are evaluated and compared on EGG data from BCI competition data
sets; results show that robust methods outperformed classical techniques for subjects with poor classification accuracy. "