Joint channel-frequency selection for motor imagery-based BCIs using a semi-supervised SVM algorithm

In this paper, motor imagery-based Brain computer interfaces (BCIs) are considered, in which channels and frequency band are two important parameters. A semi-supervised support vector machine algorithm is proposed for joint channel-frequency selection automatically and adaptively. This algorithm is designed for small training data case, in which the training data set is insufficient for parameter setting. Our algorithm is then applied to a BCI competition data set. Data analysis results are presented and the effectiveness of this algorithm is demonstrated.