Real Time Drowsiness Detection Based on Facial Dynamic Features

Abstract

This paper proposes a real-time driver drowsiness detection method based on yawning behaviors. A drowsy person usually will close their eyes and yawn for a while. We proposed five indicators: eyes aspect ratio (EAR), mouth aspect ratio (MAR), the average of eyes opening duration (EARAVEG), the average of mouth opening duration (MARAVEG), and the ratio of eyes and mouth opening duration (RATIO) as the selected features and use support vector machine SVM classifier for training and testing to improve the detection accuracy. The proposed method uses facial landmarks to model facial dynamics features of eyes and mouth while yawning. The advantages of the proposed method do not require expensive add-ons equipment and low computation cost. The experiment uses the YawDD public database to demonstrate the performance. Experimental results show the best accuracy is 96.5%, and the average processing time is 0.22 s, suggesting the proposed method is detection efficient and suitable for real-time drowsiness detection.