Driver Drowsiness Detection using Smart Glasses

Traffic accidents caused by drowsy drivers pose a significant threat to road safety, accounting for a substantial portion of the 270 million annual accidents. Addressing this issue has led to the development of various drowsiness detection methods, from physiological sensors to advanced machine learning algorithms. Despite extensive research and commercial devices, identifying the most suitable and least intrusive methods for practical use remains unclear. This paper traces the evolution of detection techniques and introduces a novel driver drowsiness system using smart glasses powered by a Raspberry Pi 4. Utilizing Dlib and a machine learning toolkit, the system analyzes facial data, employing Eye Aspect Ratio and Mouth Aspect Ratio for classification. Upon detecting the SLEEPING state, cross-referenced with monitored heart rate, the system triggers three phases of alerts. By combining multiple detection methods, the hybrid model offers improved accuracy and reliability over non-hybrid approaches, providing a more robust solution for detecting driver drowsiness.