Linear Regression Algorithm for Hand Tapping Recognition Using Functional Near Infrared Spectroscopy

Abstract

This paper proposed a linear regression (LR) algorithm for hand tapping recognition using functional Near Infrared Spectroscopy (fNIRS). Brain data with noise and artifacts were re-processed to obtain data smoothy using a Savitzky-Golay filter. The smoothy data were calculated using the proposed LR algorithm in order to produce the angular coefficients of the straight lines which correspond to oxygen-Hemoglobin (Oxy-Hb) concentration. Therefore, one can distinguish the right and left hand tapping tasks based on the different angular coefficients of the lines corresponding to the difference of the right and left brain Oxy-Hb. In addition, the difference of the left and right brain activities were determined based on comparing the angular coefficients. Experimental results showed to illustrate the effectiveness of the proposed method.