Unobtrusive drowsiness detection by neural network learning of driver steering

The purpose of this study is to detect drowsiness in drivers unobtrusively to prevent accidents and to improve safety on the highways. A method for de tecting drowsiness/sleepiness in drivers is developed. This method is based on an artificial neural network (ANN). Steering angle signals are preproce ssed and presented to the ANN which classifies them into drowsy and non-dro wsy driving intervals. The method presented here relies on signals from the vehicle steering only (steering angle) and thus presents no obstruction to the driver. A feedforward ANN was trained using an error back-propagation algorithm and tested. The training and testing data were obtained from a pr evious experiment in a driving simulator driven by 12 drivers, each under d ifferent levels of sleep deprivation. The network classifies driving interv als into drowsy and non-drowsy intervals with high accuracy.