Estimation of mental workload during automobile driving based on eye-movement measurement with a visible light camera

Abstract

The aim of this study was to investigate the effective parameters of eye movement for estimation of mental workload (MWL) during automobile driving by using a visible light camera. Twelve students (6 males and 6 females) participated in this study. The participants performed a driving task with a driving simulator and a secondary task simultaneously. The N-back task was taken as the secondary task so as to control the MWL. The levels of N-back task were as follows: none, 0-back, 1-back, 2-back, and 3-back. The gaze and head angles and blink frequency were measured with a visible light camera. The eyeball rotation angle was calculated from the gaze and head angle. The sharing rate of head movement, which is the ratio of head movement against the gaze movement, was calculated. The subjective MWL and accuracy of N-back task were also measured. The result showed that the subjective MWL monotonically increased with the increase of the difficulty of N-back task. The effect of the difficulty of N-back task was statistically significant for the standard deviations (SDs) of horizontal and vertical gaze angle, SD of horizontal eyeball rotation angle, sharing rate of head movement in horizontal direction, and blink frequency. The result of logistic regression analysis showed that the SD of horizontal eyeball rotation angle and blink frequency were the most significant parameters to estimate the MWL. The area under the curve (AUC) of receiver operating characteristic (ROC) curve for the regression equation was 0.822; thus, it has relatively high discrimination ability on the MWL.