A Visual Spelling System Using SSVEP Based Hybrid Brain Computer Interface with Video-Oculography

TLDR: A hybrid Brain Computer Interface system is developed using steady state visual evoked potential (SSVEP) along with the video-oculogram (VOG) for frequency recognition and the canonical correlation analysis (CCA) is used for SSVEP frequency recognition.

Abstract: A hybrid Brain Computer Interface (BCI) system is developed using steady state visual evoked potential (SSVEP) along with the video-oculogram (VOG). The keyboard layout is designed with 23 characters flickering at selected frequencies. The template matched webcam images provide the direction of eye gaze information to localize the user gazing space on the visual keyboard/display. This spatial localization helps to use/make multiple stimuli of the same frequency. The canonical correlation analysis (CCA) is used for SSVEP frequency recognition. The experiments were conducted on 8 subjects for both online and offline analysis. Based on the classification accuracy from offline analysis, the subject specific SSVEP stimulus duration and the optimal number of EEG channels were selected for online analysis. An average online classification accuracy of 93.5% was obtained with the information transfer rate (ITR) of 96.54 bits/min without inter character identifying delay. When a delay of 0.5 s is introduced between stimulus window the ITR of 80.17 bits/min is realized.