Improvement of Algorithm in Real-Time Brain–Computer Interface

Abstract

A brain–computer interface (BCI) using the visual evoked potential (VEP) is being investigated for the use by persons with physical disabilities because it can be input only by looking at the visual stimulus. A transient VEP (TRVEP), which is a type of VEP, is a shape analysis method for synchronous addition during blinking stimuli and was applied only for blinking stimuli below 3.5 Hz. Therefore, in our research, when TRVEP analysis was conducted for high-speed blinking stimuli of  ≥3.5 Hz, it was possible to discriminate the blinking stimulus gaze in approximately 2 s at 10 Hz. Furthermore, it has been shown that discrimination can be performed using a lighting interval fluctuation stimulus instead of a regular interval blinking stimulus. This approach currently enables the discrimination of a maximum of eight types of gazes.

In this study, we improved the algorithm in the discrimination of eight options by Electroencephalography (EEG) and investigated the discrimination rate using the correlation coefficient with the sample waveform. As a result, the discrimination rate was lower than that reported in previous studies. One of the factors is that the correlation coefficient decreased because the synchronization phenomenon did not occur. Therefore, it is conceivable to discriminate only the synchronized part instead of the correlation coefficient. There is the possibility that the judgment process can be shortened by narrowing the waveform information utilized for the judgment. In a previous study, discrimination was performed using a coefficient of variation that can simultaneously evaluate the waveform amplitude and standard deviation. These methods will also be examined. In addition, since the blinking control has been improved, we will consider using a stimulus that shifts the phase of the 10 Hz blinking stimulus.