Detection of Perceived and Imagined Rhythm from single trial ERP
Peter Desain
Univ. Nijmegen


Listening to music implies an ongoing process of confirmation and
violation of temporal expectancies. The ERP signal, measured from
electrodes on the scalp, contains traces thereof. These rhythmic
signatures can be extracted from the background-noise by appropriate
signal processing methods and allow for a direct detection. A
Bayesian analysis based on the distribution of the signals was
elaborated, not only considering their means, which is more common in
ERP work. The stochastic framework enables an easy combination of
information over time, over frequency bands and channels. This
resulted in good classification results, identifying which temporal
pattern was perceived or imagined from a small set of short musical
rhythms. But for true Brain Computer Interfacing, the robustness of
the method still needs to be enhanced further.
back to the list of talks