Determination of temporal window size for classifying the mean value of fNIRS signals from motor imagery

In this paper we classify the functional near-infrared spectroscopy (fNIRS) signals corresponding to right-and left-wrist motor imagery using various temporal windows of the response data. Signals are acquired from the primary motor cortex of five healthy subjects during right- and left-wrist motor imagery tasks using a continuous wave fNIRS system. Linear discriminant analysis is used to classify the mean values of the change in concentration of oxygenated hemoglobin with an average accuracy of 75.22%, across all subjects, for the signals acquired during the entire task period. The classification accuracies are increased to 79.82% when the analysis time is reduced by removing the initial 2 seconds of the response data. These results demonstrate the feasibility of fNIRS for a brain-computer interface.