Shrinkage estimator based common spatial pattern for multi-class motor imagery classification by hybrid classifier

Motor imagery BCI is a system that is very useful to help people with disabilities who can't move their limbs. These systems use brain activity patterns that are made from motor imagery without actual movement. In this paper, we proposed enhanced One Versus One (OVO) structure to classify EEG-based multi-class motor imagery signals. Also, shrinkage estimator based Common Spatial Pattern (CSP) is used to overcome disadvantages of conventional CSP. Shrinkage estimator is a procedure to estimate covariance matrix that regularizes CSP versus overfitting. The results of four-class classification of BCI competition IV dataset 2a, show that the performance is improved to 0.61 kappa score.