Utilizing SVMs to derive psychophysiological information from a Brain-Computer Interfacing study.

Brain-Computer Interfaces (BCIs) enable people to control Computers by thought. Mostly, BCIs utilize Electroencephalography (EEG) to “translate” cognitive processes into control commands. In this study, we combined the BCI-approach of Farwell & Donchin (1988) based on the P300-component in EEG-patterns with recent machine-learning techniques based on Support Vector Machines (SVM) with Gaussian kernels. We demonstrated, that with this combination, we can achieve very high transfer rates of 84.7 bits/min (Meinicke et al., 2003). It would therefore be nice to know, what was learned by the SVM, rather than solely using it for classification.