UGent researchers develop first self-learning Brain-Computer Interface

Tue, 08/07/2012 - 11:39

A Brain-Computer Interface (BCI) allows one to control a computer by thought. Such a system works by analyzing the user’s brain activity to discover their intended action. The P300-based BCI speller, introduced in 1988, allows patients to type simply by looking at the desired characters. Up until now, it was necessary to calibrate the P300 speller. During such a calibration session, the patient has to spell a predefined text. Analyzing the recorded brain signals in combination with the given text allows the computer to learn how the brain waves indicate the desired characters. These calibration sessions are very time-consuming and have to be repeated for every new patient. The novel system does not need any calibration. The patient can start spelling right away. The BCI system needs a very limited amount of characters to discover the underlying structure of the brain signals. After that, the system performs as well as a fully calibrated system. This breakthrough brings us another step closer to widespread use of such a Brain-Computer Interface.