A real-time non-intrusive FPGA-based drowsiness detection system

Abstract

Automotive has gained several benefits from the Ambient Intelligent researches involving the deployment of sensors and hardware devices into an intelligent environment surrounding people, meeting users’ requirements and anticipating their needs. One of the main topics in automotive is to anticipate driver needs and safety, in terms of preventing critical and dangerous events. Considering the high number of caused accidents, one of the most relevant dangerous events affecting driver and passengers safety is driver’s drowsiness and hypovigilance. This paper presents a low-intrusive, real-time driver’s drowsiness detection system for common vehicles. The proposed system exploits the ‘‘bright pupil’’ phenomenon generated by a 850 nm IR source light embedded on the car dashboard. This visual effect, due to the retina’s property of reflecting the 90% of the incident light, makes easier the detection of driver’s eyes. At the same time, the ‘‘bright pupil’’ effect is used to quantify the driver’s drowsiness level as the percentage of time in which the driver’s eyes are closed more than 80%. The efficiency of the image processing chain, together with an embedded hardware device exploiting the availability of mature reconfigurable hardware technology, such as Field Programmable Gate Array, allow to implement a real-time detection system able to process an entire 720 9 576 frame in 16.7 ms. The effectiveness of the proposed system has been successfully tested with a human subject operating in real conditions.