Driver Fatigue Detection Using Frequency‑modulated Continuous Millimeter‑wave Radar and Machine Learning with Feature Extraction

vector machineDistracted and fatigued driving is a major cause of traffic accidents worldwide, with statistics in Taiwan showing that 20% of annual accidents are attributed to these factors, ranking fatigue as the second leading cause.Drivers often experience reduced reaction times and impaired decision-making after continuous driving for 2-3 h, with post-meal and midnight driving further increasing drowsiness.Conventional fatigue detection methods, such as facial monitoring systems, can reduce accident rates by 10-20%, but they are limited by lighting conditions and raise privacy concerns.To solve these problems, a fatigue detection system is developed using frequency-modulated continuous-wave millimeter-wave radar.Using the Texas Instruments AWR1642boost development board, data were collected and analyzed through a multi-expert framework integrating four features: nodding, yawning, respiration, and heart rate.A support vector machine classifier was employed to enhance detection accuracy.The indoor simulation with 20 participants for 500 min of driving scenarios shows proof-of-concept feasibility, with an accuracy of 90.4%, a precision of 85.3%, and a recall of 88.2%.However, engine vibrations, cabin noise, road conditions, body positions, and driver fatigue must be considered in the simulation for further validation of the system.It is necessary to include larger, more diverse participant groups for on-road evaluations in complex environments.