CONF chavarriaga:iccn:2007/IDIAP To Err Is Human: Learning from Error Potentials in Brain-Computer Interfaces Chavarriaga, Ricardo Ferrez, Pierre W. Millán, José del R. EXTERNAL http://publications.idiap.ch/attachments/papers/2007/chavarriaga-iccn-2007.pdf PUBLIC http://publications.idiap.ch/index.php/publications/showcite/chavarriaga:rr07-37 Related documents 1st International Conference on Cognitive Neurodynamics (ICCN 2007) 2007 IDIAP-RR 07-37 Several studies describe evoked EEG potentials elicited when a subject is aware of an erroneous decision either taken by him or by an external interface. This paper studies {\em Error-related potentials} (ErrP) elicited when a human user monitors an external system upon which he has no control whatsoever. In addition, the possibility of using the ErrPs as a learning signals to infer the user's intended strategy is also addressed. Experimental results show that single-trial recognition of correct and error trials can be achieved, allowing the fast learning of the user's strategy. These results may constitute the basis of a new kind of human-computer interaction where the former provides monitoring signals that can be used to modify the performance of the latter.This work has been supported by the Swiss National Science Foundation NCCR-IM2 and by the EC-contract number BACS FP6-IST-027140. This paper only reflects the authors' views and funding agencies are not liable for any use that may be made of the information contained herein. REPORT chavarriaga:rr07-37/IDIAP To Err Is Human: Learning from Error Potentials in Brain-Computer Interfaces Chavarriaga, Ricardo Ferrez, Pierre W. Millán, José del R. EXTERNAL http://publications.idiap.ch/attachments/reports/2007/chavarriaga-idiap-rr-07-37.pdf PUBLIC Idiap-RR-37-2007 2007 IDIAP Several studies describe evoked EEG potentials elicited when a subject is aware of an erroneous decision either taken by him or by an external interface. This paper studies {\em Error-related potentials} (ErrP) elicited when a human user monitors an external system upon which he has no control whatsoever. In addition, the possibility of using the ErrPs as a learning signals to infer the user's intended strategy is also addressed. Experimental results show that single-trial recognition of correct and error trials can be achieved, allowing the fast learning of the user's strategy. These results may constitute the basis of a new kind of human-computer interaction where the former provides monitoring signals that can be used to modify the performance of the latter.This work has been supported by the Swiss National Science Foundation NCCR-IM2 and by the EC-contract number BACS FP6-IST-027140. This paper only reflects the authors' views and funding agencies are not liable for any use that may be made of the information contained herein.