Asynchronous detection of error-related potentials using a generic classifier

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

Error-related potentials (ErrPs) can be usedto improve BCIs’ performance but its use is often with-held by long calibration periods. We recorded EEG dataof 15 participants while controlling a robotic arm towardsa target. In 30 % of the trials, the protocol prompted anerror during the trial in order to trigger ErrPs in the partic-ipants. For each participant, we trained an ErrP classifierusing the data of the remaining 14 participants. Each ofthese classifiers was tested asynchronously on the data ofthe selected participant. The threshold that maximizedthe product of the average true positive rate (TPR) andthe average true negative rate (TNR) wasτ=0.7. For thisthreshold, the average TPR was 53.6 % and the averageTNR was 82.0 %. These results hint at the feasibility oftransferring ErrPs between participants as a reliable strat-egy to reduce or even remove the calibration period whentraining ErrP classifiers to be used in an asynchronous manner
Original languageEnglish
Title of host publicationProceedings of the 8th Graz Brain Computer Interface Conference 2019
Subtitle of host publicationBridging Science and Application
EditorsGernot R. Müller-Putz, Jonas C. Ditz, Selina Wriessnegger
Place of PublicationGraz
PublisherVerlag der Technischen Universität Graz
Pages54-58
Number of pages5
ISBN (Electronic)978-3-85125-682-6
DOIs
Publication statusPublished - 2019
Event8th Graz Brain-Computer Interface Conference 2019: Bridging Science and Application - Petersgasse 16, Graz, Austria
Duration: 16 Sep 201920 Sep 2019
Conference number: 8
https://www.tugraz.at/institutes/ine/graz-bci-conferences/8th-graz-bci-conference-2019/

Conference

Conference8th Graz Brain-Computer Interface Conference 2019
Abbreviated titleGBCIC 2019
CountryAustria
CityGraz
Period16/09/1920/09/19
Internet address

Fingerprint Dive into the research topics of 'Asynchronous detection of error-related potentials using a generic classifier'. Together they form a unique fingerprint.

Cite this