부산대학교 도서관

Electroencephalographic signals reflect a superposition of activity from neural populations both related and unrelated to the brain processes under investigation. Here we show how these complex signals, containing a considerable amount of noise, can be decoded to send binary messages only by covertly directing attention to a visual target stimulus. We used canonical correlation analysis to decompose the characteristic series of event-related potentials that correlated with a given sequence of stimuli comprising the target. Across 13 participants, we were able to correctly decode the responses to more than 90% of dichotomous questions by using our brain–computer interface (BCI). Although the system relies on visual stimuli, it neither depends on eye movements nor on high visual acuity. Therefore, this BCI might be suitable for patients who are unable to communicate due to motoneuron disorders resulting in complete paralysis of all voluntarily-controlled musculature.