Novel Characterization of the Steady-State Visual Evoked Potential Spectrum of EEG

Steady-State Visual Evoked Potentials (SSVEPs) are oscillations of the electroencephalogram (EEG) observed over the occipital area that exhibit a frequency corresponding to a repetitively flashing visual stimulus. SSVEPs have proven to be very consistent signals for rapid EEG-based braincomputer interface (BCI) control. However, due in part to perceptual and neurophysiological aspects, SSVEP signal detection biases exist for different stimulation frequencies. Furthermore, these biases tend to differ across subjects. Canonical correlation analysis (CCA) has proven to be the most robust approach for detecting SSVEPs in multiclass stimulus paradigms where each potential target flashes at a different frequency. In this work, in order to provide a better characterization of the SSVEP spectrum for BCI applications, 22 subjects were stimulated with an LED array that flashed according to a chirp signal having a frequency that varied over the typical functional range of SSVEP from 5.5-34.5 Hz. The resulting EEG was analyzed using CCA to elucidate the stimulus frequencies that produce the best discriminability for practical use. Subjects achieved an average accuracy of 72.2% using a six-class paradigm with a standardized set of stimulus frequencies. However, when using a subject-specific frequency set (i.e. frequencies optimized for each subject), the average accuracy significantly increased to 83.7% (p = 0.03). The results show that inherent SSVEP response differences exist between subjects, which can have a significant effect on performance. This approach also establishes a framework for a rapid optimization of subject-specific frequency profiles.