Reduction in Command Generation Time for fNIRS-Based BCI

• Published in: Asian Control Conference (Online) pp. 36 - 41
• Main Authors: Khan, M. N. Afzal, Ghafoor, Usman, Yang, Dalin, Hong, Keum-Shik
• Format: Conference Proceeding
• Language: English
• Published: ACA 04.05.2022
• Subjects: Band-pass filters; Brain; Brain-computer interface (BCI); classification; Data mining; Feature extraction; Fingers; fNIRS; Information filters; short stimulation; Visualization
• ISSN: 2770-8373
• DOI: 10.23919/ASCC56756.2022.9828043

Stimulating brain in order to generate certain brain activity is one of the core research issues in the field of brain-computer interface (BCI). One of the major debatable topics is the duration of that stimulation. To comment on this issue, two different stimulation durations in sensorimotor and visual cortices of the brain are investigated in this study. In order to acquire brain signals for three different tasks performed (right hand index finger poking, right hand index finger tapping, and checkboard) functional near-infrared spectroscopy (fNIRS) was used. This study tried to investigate the classification accuracy that can be achieved using small stimulation durations i.e., 1 sec and 3 sec. The influence of these two stimulation durations on the hemodynamic response (HR) signal was monitored using fNIRS. In total six participants took part in this study. Simultaneous monitoring of sensorimotor and visual cortices of all the subjects were done during the experiments. The acquired data was first preprocessed and filtered using band-pass filter. The filtered data was then used to extract different features. Different temporal characteristics and numerous time frames were used to extract temporal information after stimulation. The extracted features were used to classify the three different tasks for both of the stimulation durations. The results of the study suggested that brain command for fNIRS-based BCI can be generated within 1~3 sec with accuracies much high above the chance level. These results provide the evidence that long stimulation durations can be avoided and command for BCI can be generated using 1~3 sec stimulation durations.