Vehicle Driver State Estimation Using AI and Algorithm

Drowsy driving is the leading cause of road accidents, which leads to the development of real-time driver monitoring systems. Traditional methods, like tracking vehicle movements or using wearable sensors, can be unreliable or uncomfortable for drivers. This paper suggests a non-invasive, continuous driver monitoring system combining computer vision-based drowsiness detection with remote Photoplethysmography (rPPG) for heart rate estimation. The system makes use of a webcam to identify eye closure with Eye Aspect Ratio (EAR) and also estimate heart rate through examination of minor variations in skin color on the forehead area. The system also tracks head motion (yaw angle) to identify distractions. Upon detection of drowsiness or unusual physiological activity, an audio warning is initiated to alert the driver. The system was tested on 175 subjects, including 50 real-time participants and 125 from open-source datasets. Our experiments show the system is highly accurate, with a 91.42% success rate in detecting drowsiness and 84.21% accuracy in heart rate estimation under normal lighting. The proposed solution is an inexpensive and scalable option for promoting the safety of drivers and preventing fatigue-related crashes.