The technology is based on steering wheel movements - which are more variable in drowsy drivers - and offers an affordable and more reliable alternative to currently available video-based driver drowsiness detection systems, researchers said.
"Video-based systems that use cameras to detect when a car is drifting out of its lane are cumbersome and expensive," said Hans Van Dongen, research professor at the Washington State University Sleep and Performance Research Center.
"They don't work well on snow-covered or curvy roads, in darkness or when lane markers are faded or missing.
In an experiment, 29 participants were on a simulated 10-day night shift schedule that caused moderate levels of fatigue, as assessed by their performance on a widely used alertness test known as the psychomotor vigilance task (PVT).
Also Read
During each night shift, participants spent four 30-minute sessions on a high-fidelity driving simulator, which captured data for 87 different metrics related to speed, acceleration, steering, lane position and other factors.
Researchers then showed that data on steering wheel variability can be used to predict variability in lane position early on, making it possible to detect driver drowsiness before the car drifts out of its lane.
"We wanted to find out whether there may be a better technique for measuring driver drowsiness before fatigue levels are critical and a crash is imminent," Van Dongen said.
"Our invention provides a solid basis for the development of an early detection system for moderate driver drowsiness. It could also be combined with existing systems to extend their functionality in detecting severe driver drowsiness," he said.
The science behind the invention was described in the journal Accident Analysis & Prevention.