Calibration-free SSVEP-based BCI Switch

Steady state visually evoked potentials (SSVEP) are periodic responses elicited when a participant looks at a visual stimulus flickering at a particular frequency. Brain computer interfaces (BCI) based on SSVEP have been shown to have consistent responses across most human subjects. We use this property for designing a two-state BCI switch. Canonical correlation analysis (CCA) is used to distinguish between the intentional control and the no control states of the switch. The traditional approach of using CCA for SSVEP detection uses the target frequencies only in the reference templates. This cannot be directly used to realize a BCI switch, since the templates do not incorporate the background EEG activity. To overcome this, we propose a template frequency selection method. The study is carried out on seven healthy participants. The proposed method achieves an average F1-Score of 0.95 (offline) and 0.85 (online). The mean activation time is found to be of 2.09 s (offline) and 3.01 s (online) across participants. These results indicate that the proposed system can be effectively used as a two-state BCI switch to enable control of external applications.