Small Sample Setting and Frequency Band Selection Problem Solving Using Subband Regularized Common Spatial Pattern

Abstract

For the last few years, motor imagery electroencephalogram (EEG)-based brain-computer interface (BCI) systems have received the significant amount of attention in various areas, including medicine and engineering. The common spatial pattern (CSP) algorithm is the most commonly used method to extract features from motor imagery EEG. However, the CSP algorithm has limited applicability in small-sample setting (SSS) situations because these rely on a covariance matrix. In addition, the large differences in performance depend on the frequency bands that are being used. To address these problems, 4–40 Hz band EEG signals are divided into nine subbands, and regularized CSP (R-CSP) is applied to individual subbands. Fisher’s linear discriminant (FLD) is applied to the features of R-CSP extracted from individual subbands, and the results obtained through the foregoing are connected for all subbands to make an FLD score vector. Furthermore, principal component analysis is applied to use the FLD score vectors as the inputs of the classifier least square support vector machine. The proposed method yielded a classification accuracy of 86.61%, 98.21%, 63.78%, 87.05%, and 77.78% from five subjects (“ $aa$ ”, “ $l$ ”, “ $av$ ”, “ $aw$ ”, and “ $ay$ ”, respectively) for BCI competition III data set IVa by using 18 channels in the vicinity of the motor area of the cerebral cortex. The proposed method offers particularly excellent performance in SSS situations.