Abstract
When inputting letters in non-alphabetic languages, there are two input interfaces: roman input or input language letters. When inputting Japanese letters, the Japanese fifty-letter type interface is more efficient than the alphabetic interface. Among the interfaces that use EEG to input letters, the one that uses steady state visual evoked potential (SSVEP), one of visual evoked potential, is called SSVEP-brain-computer interface (BCI). The purpose of this study is to design and evaluate SSVEP-BCI using Japanese fifty-letter type, which is more efficient than Roman letter input using alphabetic table. In order to handle 50 different input types in SSVEP-BCI, we proposed methods such as stimulus frequency design and fusion of display space and analysis algorithms. In particular, the analysis method using the positional relationship in the display space for SSVEP-BCI contains many novelties. As a result, we achieved an accuracy of 77.10% and an ITR of 75.08 bits/min. This is equivalent to 15.42 Japanese fifty-letter type per minute. We also evaluated the positional relationship between input and output objects in the display space. The study showed that there were many misjudgments due to the selection of horizontally adjacent objects in the display.
