Classification of Steady-State Visual Evoked Potentials based on the visual stimuli duty cycle

The detection of Steady State Visual Evoked Potentials (SSVEP) in the electroencephalogram (EEG) allows creating non-invasive Brain-Computer Interface (BCI). To produce an SSVEP response, a visual stimulus must be presented to the user. This stimulus can be a light that flickers at a particular frequency. Classical SSVEP-BCIs consider a frequency for each BCI command. One problem for an SSVEP based BCI can be the number of simultaneous flickering stimuli. It is difficult to render many flashing boxes with as many frequencies as boxes, due to hardware constraint like the vertical refresh rate of a screen. As an alternative to the common paradigm that assigns one command to each frequency, we propose to classify different type of SSVEP responses based on the duty cycle of the flickering lights, the frequency being the same for evoking SSVEP responses. Three paradigms based on different duty cycles over six subjects are compared. The offline classification of the obtained SSVEP responses is performed with spatial filters combined with a Bayesian Linear Discriminant Analysis classifier. The results show that it is possible to efficiently discriminate SSVEP responses given by visual stimuli at the same frequency but with different duty cycles.