Automatic determination of the optimum stimulation frequencies in an SSVEP based BCI

A brain computer interface (or BCI) is a communication or control system which does not need any neuromuscular activity to produce a message or action. In this project, an SSVEP based BCI is used. This means that when the user concentrates on a LED, flickering with a certain frequency, the steady state visual evoked potential that is caused by this flickering and measured with EEG electrode Oz, will control the system. The amplitude of the SSVEP is expected to be different in each subject. This inter subject difference motivates for BCI calibration, since a higher SSVEP amplitude implies a better performance of the BCI. The inter subject difference is investigated in this project. Further, for calibration, it is convenient to use a short (<10 min) calibration sequence, with which the SSVEP amplitude is predicted. System identification techniques are used in this process. In this project, linear and nonlinear approaches are used to predict the SSVEP. The reliability of this prediction is investigated. Conclusions: • The SSVEP is measured in four subjects, inter subject difference was shown. • Linear approaches did not suffice in reliably predicting the SSVEP amplitude. • There were no nonlinear approaches found that reliably predict the SSVEP amplitude.