Bayesian Classification of Single-Trial Event-Related Potentials in EEG
Jens Kohlmorgen and Benjamin Blankertz
International Journal of Bifurcation and Chaos
We present a systematic and straightforward approach to the problem of single-trial classi cation of event-related potentials (ERP) in EEG. Instead of using a generic classi er o -the-shelf, like a neural network or support vector machine, our classi er design is guided by prior knowledge about the problem and statistical properties found in the data. In particular, we exploit the well-known fact that event-related drifts in EEG potentials, albeit hard to detect in a single trial, can well be observed if averaged over a su ciently large number of trials. We propose to use the average signal and its variance as a generative model for each event class and use Bayes' decision rule for the classi cation of new, unlabeled data. The method is successfully applied to a data set from the NIPS*2001 Brain-Computer Interface post-workshop competition. Our result turns out to be competitive with the best result of the competition.