de.mpg.escidoc.pubman.appbase.FacesBean
English
 
Help Guide Disclaimer Contact us Login
  Advanced SearchBrowse

Item

ITEM ACTIONSEXPORT

Released

Poster

EEG Channel Selection for Brain Computer Interface Systems Based on Support Vector Methods

MPS-Authors
http://pubman.mpdl.mpg.de/cone/persons/resource/persons84916

Schröder,  M
Max Planck Institute for Biological Cybernetics, Max Planck Society;

http://pubman.mpdl.mpg.de/cone/persons/resource/persons84035

Lal,  TN
Department Empirical Inference, Max Planck Institute for Biological Cybernetics, Max Planck Society;

http://pubman.mpdl.mpg.de/cone/persons/resource/persons84193

Schölkopf,  B
Department Empirical Inference, Max Planck Institute for Biological Cybernetics, Max Planck Society;

Locator
There are no locators available
Fulltext (public)
There are no public fulltexts available
Supplementary Material (public)
There is no public supplementary material available
Citation

Schröder, M., Lal, T., Bogdan, M., & Schölkopf, B. (2004). EEG Channel Selection for Brain Computer Interface Systems Based on Support Vector Methods. Poster presented at 7th Tübingen Perception Conference (TWK 2004), Tübingen, Germany.


Cite as: http://hdl.handle.net/11858/00-001M-0000-0013-D9E5-7
Abstract
A Brain Computer Interface (BCI) system allows the direct interpretation of brain activity patterns (e.g. EEG signals) by a computer. Typical BCI applications comprise spelling aids or environmental control systems supporting paralyzed patients that have lost motor control completely. The design of an EEG based BCI system requires good answers for the problem of selecting useful features during the performance of a mental task as well as for the problem of classifying these features. For the special case of choosing appropriate EEG channels from several available channels, we propose the application of variants of the Support Vector Machine (SVM) for both problems. Although these algorithms do not rely on prior knowledge they can provide more accurate solutions than standard lter methods [1] for feature selection which usually incorporate prior knowledge about neural activity patterns during the performed mental tasks. For judging the importance of features we introduce a new relevance measure and apply it to EEG channels. Although we base the relevance measure for this purpose on the previously introduced algorithms, it does in general not depend on specic algorithms but can be derived using arbitrary combinations of feature selectors and classifiers.