A new AI algorithm can detect if a driver is drunk with 75% accuracy by constantly monitoring their facial features. This technology, integrated into in-vehicle cameras, could significantly reduce drunk driving accidents by identifying intoxication levels as soon as the driver gets in.
Researchers say this system surpasses existing methods that rely on data collected only while the vehicle is moving, such as steering patterns, pedal usage, and vehicle speed. Instead, it uses a single color camera to observe variables like gaze direction and head position, and can also incorporate 3D and infrared footage of the driver’s face and rearview videos showing driver posture.
The project, detailed in a paper published on April 9 at an IEEE and CVF conference, highlights the potential of in-car computing systems to prevent impaired driving. Ensiyeh Keshtkaran, a doctoral student at Edith Cowan University, Australia, noted that the software fits seamlessly into smart vehicle architectures, making it easy to adapt to other environments like smartphones.
According to the World Health Organization, alcohol impairment is involved in 20% to 30% of fatal car accidents worldwide. In Australia, where this project originated, 30% of fatal crashes involve blood alcohol levels over the legal limit of 0.05%.
The study used video footage of drivers with various ages, drinking habits, and driving experience in simulators under three levels of intoxication: sober, low intoxication, and severely intoxicated. Partnering with MiX by Powerfleet, researchers collected data from alcohol-impaired drivers in controlled but realistic environments. The algorithm identified intoxication by recognizing facial cues like bloodshot eyes, flushed face, droopy eyelids, and a dazed look, correctly predicting a driver’s state in 75% of cases.
Project lead Syed Zulqarnain Gilani, senior lecturer at Edith Cowan University, aims to improve the resolution of the image data to enhance prediction accuracy. This advancement could lead to a future where smart cars prevent drunk drivers from starting the vehicle or even alert authorities if necessary.
