Decoding fNIRS based imagined movements associated with speed and force for a brain-computer interface

Abstract

Functional near-infrared spectroscopy (fNIRS) is an emerging non-invasive technology applied in brain-computer interface (BCI). This study investigates fNIRS based imagined hand-clenching tasks, indicating that the combinations of speed and force have distinct patterns which can be decoded to develop a BCI system. Twelve healthy participants are instructed to perform imagined left or right hand-clenching tasks; oxy-hemoglobin (HbO2) and deoxy-hemoglobin (Hb) concentrations are acquired from motor cortex using a multi-channel fNIRS system. Feature selection method based on mutual information is employed to select the optimal features for classification, and support vector machine (SVM) is used as a classifier resulting in average accuracies of 84.9% and 86.1% for classifying left and right imagined movements. Compared with traditional fNIRS-BCI system, this study provides a possibility to generate a new control pattern for brain-controlled robots, e.g., speed or force control. There is a potential application to combine fNIRS-BCI system with exoskeleton for rehabilitation.