Driver yawning detection based on long short term memory networks

This paper proposes an efficient and nonintrusive approach for detecting Driver's Yawn using a single camera based on Long Short Term Memory Networks. Fatigue is a significant cause of accidents since it reduces drivers' capacity of responding to emergency. Because drivers' drowsiness involves several external facial expression including continuous yawns. We propose an system employing Convolutional Neural Networks to extract spatial image features and an Long Short Term Memory Network to Analyze temporal characteristics. Experimental results indicate that the proposed method perform a high accuracy rate more than 87%. These experiments are applied over real video sequences acquired by our angle fixed web camera and the yawning detection dataset.