抄録
Many researchers have reported their studies regarding improvement of identification accuracy of the character input by electroencephalography(EEG). However, in order to develop classifier with higher identification accuracy, long time EEG measurement was required to obtain training data for machine learning before actually using interface. Therefore, in this research, by optimizing presentation of visual stimuli, we aim to identify the character inputted by comparatively simple classifier. This will shorten measurement time for the learning or skip it.
In this experiment, we tried to shorten time until character identification by optimizing visual stimulus series for flashing character arranged in the 3×3 matrix. We propose the new visual stimulus method with considering Hamming distance in order to reduce the number of flashing character at the same time and an interval between the two flashings of target character in order to increase amplitude of P300. We evaluated our method for three subjects.
Comparing our method with P300 speller which is general stimulation method, we obtained the results that our method indicated the performance of about 1.5 characters per minute more than P300 speller.
