Abstract
Brain Computer Interface (BCI) has been studied for wide variety of applications. BCI detects human brain activities as cranial nerve information and uses the information as inputs to control machine or equipment. In some studies on BCI, Near Infrared Spectroscopy (NIRS) has been used for detecting human brain activities. NIRS gives fewer restrictions on body movement than fMRI and is less affected by electric noise than EEG. Thus, NIRS is suitable for BCI.
This paper describes the development of a portable NIRS-BCI system. We used a neural network for classification of brain activities. We used moving-averaged oxygenated hemoglobin (oxy-Hb) signals and their derivatives as the feature values. The neural network was optimized using ideal NIRS signals first. Then, we investigated their classification performance using actual data recorded in an experiment with 10 participants. Finally, a portable NIRS-BCI system was constructed with the developed classifier to show the capability of real-time operation of a robot. Experimental results showed that the portable NIRS-BCI system gives the real-time operation of the robot by the user's brain activity.
