Unsupervised Learning in Brain-Computer Interfaces: Theory and Practice
09:00 - 12:00, 28 January

1/2 day
Beginner level

@ Room 5BC


Brain-Computer Interfaces (BCI) can translate brain signals into control commands, e.g. allowing paralysed patients to communicate or control a wheelchair. The measured signals strongly vary from user to user and day to day. Thus, a daily calibration session is normally needed. In the workshop, we explain the basics of Brain-Computer Interfaces and discuss how recent unsupervised adaptive machine learning methods can reduce or even completely eliminate the need for a calibration session, hence drastically increasing the usability of these systems. In a live demo, we show how a subject can spell a sentence without any calibration by using only his brain activity measured by electroencephalography (EEG).


The participants should get familiar with the basics of Brain-Computer Interfaces (How do you prepare an EEG-cap? Which signals are acquired by the EEG? How are these signals processed?). Additionally, participants should understand how unsupervised adaptive learning can be used in Brain-Computer Interfaces.


David Hübner

PhD Student, University of Freiburg

Website   ·   Email

Michael Tangermann

Group Leader of the Brain State Decoding Lab, University of Freiburg