COMPARISON OF FOUR CLASSIFICATION METHODS FOR BRAIN-COMPUTER INTERFACE

Abstract

This paper examines the performance of four classiflers for Brain Com- puter Interface (BCI) systems based on multichannel EEG recordings. The clas- siflers are designed to distinguish EEG patterns corresponding to performance of several mental tasks. The flrst one is the basic Bayesian classifler (BC) which exploits only interchannel covariance matrices corresponding to difierent mental tasks. The second classifler is also based on Bayesian approach but it takes into account EEG frequency structure by exploiting interchannel covariance matrices estimated separately for several frequency bands (Multiband Bayesian Classifler, MBBC). The third one is based on the method of Multiclass Common Spatial Pat- terns (MSCP) exploiting only interchannel covariance matrices as BC. The fourth one is based on the Common Tensor Discriminant Analysis (CTDA), which is a generalization of MCSP, taking EEG frequency structure into account. The MBBC and CTDA classiflers are shown to perform signiflcantly better than the two other methods. Computational complexity of the four methods is estimated. It is shown that for all classiflers the increase in the classifying quality is always accompanied by a signiflcant increase of computational complexity.