Single trial EEG classification of lower-limb movements using improved regularized common spatial pattern

Brain computer interface (BCI) is a direct communication pathway between the human central nervous system and external devices primarily aiming at restoring damaged functions such as sight, hearing and movement. Although great achievements have been made for the development of reliable BCI systems to assist people with upper-limb disabilities, researches on BCI development related to lower-limb are still rudimentary. In the current study, based on the regularized common spatial pattern analysis (R-CSP) method and statistical dependency, we have developed an improved feature selection method for lower-limb movement pattern classification. High-resolution electroencephalogram (EEG) signals were recorded from four healthy male subjects undergoing real lower-limb movements. Compared to the conventional CSP, R-CSP, and PCA methods, the proposed method achieved the best average accuracy of 83.5% for single trial classification of left and right lower-limb movement. Our findings thereby have insightful implications for developing practical BCI systems for lower-limb movement.