Deep Learning for Driver Drowsiness Classification for safe vehicle application

Drowsiness detection systems are intended to warn the drivers before increasing fatigue in order to prevent accidents. The difficulty of classifying driver vigilance in an accurate, robust, and predictive manner is a delicate task. Deep learning using different data sources, such as vehicle-based data (steering angle, mid-lane deviation, etc.), facial data (eyelid movement), and biosignals (heart rate) offer the highest potential. The chapter will summarize the different methods using deep learning and the related results in achieving accuracy, robustness and prediction. It also highlights the difficulties in obtaining signals from various data sources, pre-processing them, and finding an adequate deep learning method.