Proposal of Recognition Algorithm for Menu Selection using Steady State Visual Evoked Potential

Abstract

This study refers to a recognition algorithm for SSVEP （Steady State Visual Evoked Potential）-based brain computer interface for menu selection. Aimed at simple-to-use interfaces with no training, we propose an algorithm that requires neither calibration processes, nor machine learning. A commercial headset, “Emotiv EEG Neuroheadset,” was introduced to measure electroencephalograms in order to decrease the time required to attach electrodes. Visual stimulation with a frequency higher than 20 Hz was used to reduce the risk of photosensitive epileptic seizure. In the experiment, eight healthy young adults participated, and were asked to observe one of six squares with intensities modulated sinusoidally with six different frequencies between 20 to 30 Hz presented on an LCD （Liquid Cristal Display） with a refresh rate of 120 Hz. After studying the collected data, we have proposed an algorithm that compares the Z-scores of averaged SSVEP amplitudes of the left and right occipital lobe (O₁, O₂) obtained by the averaging method for each frequency, and selects the significantly large one. Five out of eight participants showed good performance; the average accuracy and ITR （Information Transfer Rate） are 83.3％ and 30.5 bits/min, respectively. The average number of trials required for decisions was less than two. The best two showed 100％ accuracy, and the average ITR was 35.8 bits/min.