The purpose of this study was to evaluate commercially available devices for driver fatigue monitoring with particular focus on the needs of the mining industry. We present an overview of fatigue monitoring technologies (FMTs) and propose means to evaluate the devices. Three video-based devices were selected and used in an overnight driving simulation study to test their accuracy. In total 14 healthy volunteers performed the driving simulation tasks in eight test runs separated by breaks of approximately 10 min. EEG and EOG were recorded during the driving periods. The output variable of the FMT devices (percentage of eye closures, PERCLOS), subjectively rated fatigue on the Karolinska sleepiness scale (KSS), and driving performance in terms of standard deviation of lateral position in lane (SDL) were also recorded throughout testing sessions. Regression analysis revealed that PERCLOS is significantly related to higher KSS scores and to SDL. Calculations at a finer temporal resolution as well as on an intra-subject level showed decreased correlation coefficients. Discriminant analysis of PERCLOS and EEG/EOG suggested that PERCLOS does not differentiate well between mild and strong fatigue. The results suggest that under laboratory conditions current FMT devices are reliable when temporal resolution is not too fine (>30 min) and data averaged across several subjects is utilized, but fail to give a valid prediction of subjective fatigue as well as of driving performance on an individual level. (Author/publisher)
Abstract