Sayed, R., Eskandarian, A., and Oskard, M. (2001) "Driver Drowsiness Detection Using Artificial Neural Networks", Washington, DC: Transportation Research Board.

Abstract: This paper presents a method for detecting drowsiness/sleepiness in drivers. This method is based on an Artificial Neural Network (ANN). The ANN observes the steering angle patterns and classifies them into drowsy and non-drowsy driving intervals. Most of the Driver drowsiness detection systems are based on the analysis of physical and/or physiological signals from the driver’s body such as EEG data, Eye movements, Driver body movements, and body conditions like Heart rate, Blood pressure and Body Temperature. Sensing these signals require devices that are highly obtrusive and unreliable under certain environmental conditions. We present a drowsiness detection method, which rely on signals from the vehicle only and thus present no obstruction to the driver. An ANN was trained and tested. The training and testing data was obtained from a previous experiment in a driving simulator driven by twelve drivers, each under different levels of sleep deprivation. The network classifies driving intervals into drowsy and non-drowsy intervals with high accuracy.