Machine Learning and Gradient Statistics Based Real-Time Driver Drowsiness Detection

In this paper, the machine learning and gradient statistics based driver drowsiness detection is developed for the real-time application. The proposed system includes four parts, which are the face detection, the eye-glasses bridge detection, the eye detection, and the eye closure detection. The system uses grayscale images without any color information, and it works effectively in daytime and nighttime. For the face detection, the system uses the machine learning to detect face position and face size, and the face geometrical position is used to reduce the searching range of eyes. Next, the proposed eye detection algorithm for the eye location is separated into two different modes to judge whether the driver wears glasses or not. Finally, the system detects driver's eye state in the eye region. If the driver closes their eyes during an enough time, does not concentrate on driving, or nods his head, the system generates an alarm to notify the driver. In experimental results, the average processing frame rates are up to 245 fps in a PC (i7, 2.59GHz). The average detection rate of eye closure is 91.49% when the driver wears glasses, and the corresponding detection rate is 95% when the driver does not wear glasses.