A hybrid approach for fatigue detection and quantification

Lane departure warning system, Adaptive cruise control, and Pedestrian detection are widely used in automobile systems. The developing technologies to understand and solve the problem of fatigue are quite challenging in real time automated systems. Building dependable system is difficult as we have to build road-sense into such systems, which is quite trivial for human cognitive system but difficult for an automated system. In this paper, we have proposed an unsupervised hybrid approach to quantify driver's fatigue level. We have implemented the fatigue monitoring system using modified Viola Jones algorithm, Ada-boost training method and template matching using correlation coefficient. Unintentional lane departures are also due to driver's fatigue. So we propose a non-invasive hybrid method referred to as FQS (Fatigue Quantifying System) that integrates fatigue detection system along with lane departure feature. This hybrid system aims to detect the correlation between lane departures and driver fatigue levels. Our solution combines both visual and road features to detect the drowsiness of driver to achieve more precision and reduce the false alarm rate. Preliminary results show that the proposed system has high accuracy of 80% and can quantify the fatigue levels very effectively.