A hybrid approach to feature subset selection for brain-computer interface design

Abstract

In brain-computer interface (BCI) development, temporal/spectral/ spatial/statistical features can be extracted from multiple electro- encephalography (EEG) signals and the number of features available could be up to thousands. Therefore, feature subset selection is an important and challenging problem in BCI design. Sequential forward floating search (SFFS) has been well recognized as one of the best feature selection methods. This paper proposes a filter-dominating hybrid SFFS method, aiming at high efficiency and insignificant accuracy sacrifice for high-dimensional feature subset selection. Experiments with this new hybrid approach have been conducted on BCI feature data, in which both linear and nonlinear classifiers as wrappers and Davies-Bouldin index and mutual information based index as filters are alternatively used to evaluate potential feature subsets. Experimental results have demonstrated the advantages and usefulness of the proposed method in high-dimensional feature subset selection for BCI design. © 2011 Springer-Verlag.