Abstract
Currently, it has been attracting attention for the technology of Brain-Computer Interface which reads the human thought by connecting brain to computer. Previous studies have such problems that subjects had to perform experiments in huge equipment, or to spend much time for attaching a lot of electrodes, therefore subjects had to be imposed burden. In order to improve those problems, Maruyama and Kikuchi attempted to construct simplified BCI with 2ch NIRS using two kinds of stimuli, and classified them. This study added two categories of stimuli for the study by Maruyama and Kikuchi resulting four categories, and investigated whether the classification of four categories is possible or not from the data obtained from 2ch NIRS. In our experiment, four directions of arrows were presented and the data obtained from subsequent brain activity were classified by random forest to obtain the classification rate. As a result, the classification rate over 80% on average was obtained. The result suggested that it is possible to identify the for classes even 2ch NIRS data.