Classification of a single channels fNIRS signal for a brain computer interface

Near-infrared spectroscopy (NIRS) is a state-of-the-art non-invasive modality that measures changes in hemoglobin concentrations during tasks. More and more researches tried using NIRS as signal acquisition tool for brain-computer interface (BCI). In this paper, we conducted a study to test the feasibility of using single channel fNIRS in the building of a BCI system and whether it is possible to achieve relatively high classification accuracy in motor imagery tasks from forehead channels in fNIRS-based BCI. The classification accuracy using single channel fNIRS are all higher than 73.2%, with the highest at 86.2%. It is therefore possible to use one channel to build a BCI system. From an analysis of variance on forehead and motor cortex channels classification accuracy data, no significant differences were found at p=0.05. This proves using one-channel forehead signal building fNIRS-based BCI is feasible.