Comparsion of methods for a motor imagery-based two-state self-paced brain-computer interface

Self-paced brain-computer interface (SP-BCI) has been considered a more practical BCI for users with motor disabilities. Previously, various methods have been proposed to improve the performance of the SP-BCI. However, no studies have been presented to compare the existing methods. In this study, we concentrate on a motor imagery-based two-state SP-BCI and compare the state-of-the-art methods, including various feature extraction and classification methods. Comparisons were carried on EEG data collected from four participants. True positive and false positive rates were used as the performance indices to evaluate the methods for the motor imagery-based SP-BCI. Comparison results indicate that common spatial pattern (CSP) is the best representation method. Also, support vector machine (SVM) achieves the highest true positive rate. However, its false positive rate is also the highest in most cases. Moreover, imbalanced SVM (ISVM) can keep the false positive rate below a desired threshold.