Training Session Duration Analysis of a Brain-Computer Interface

A brain-computer interface (BCI) allows patients with reduced motor abilities to interact with a computer using recordings of the brain’s electrical activity. One such method of recording is electroencephalography (EEG), a method commonly used in BCI research. A BCI is trained by a subject in often tedious sessions. In this study, the possibility of reducing the length of these sessions was investigated. Support vector machines (SVM) were trained with increasingly large parts of the session data sets and the classification accuracy was analyzed. Results showed patterns of peaks and stabilization in performance at similar points in time for different subjects. This suggests the possibility of either shortening the overall session length or tailoring session length to fit each subject, depending on the nature of the experiment