Space, Physics, and Math

The truth is in the eyes

Sleep detection technologies still lack the most accurate indicator of drowsiness

December 13, 2011

Heavy eyelids. Blurred vision. A startling jerk awake, but it’s not long before your eyes start to slowly droop again. We all know the scary feeling of drowsy driving — but can our cars accurately sense the moments before we fall asleep now, too?

About 100,000 sleep-related accidents are reported in the U.S. each year. To decrease the chances of those accidents, sleep experts agree that car companies should install drowsiness detection systems in their vehicles. Yet even the most advanced systems in cars from manufacturers such as Mercedes-Benz, Saab and Volvo cannot assess the most accurate indicator of impending sleep: our eyes.

Absent this distinct physiological cue, some experts question the ability of current technologies to accurately determine sleepiness and, perhaps more importantly, to ensure that a driver will act upon a warning.

“The concept of having a car that helps prevent you from falling asleep is a great one,” said Dr. Richard Schwab, co-director of the Penn Sleep Center at the University of Pennsylvania Medical Center. “But Mercedes hasn’t provided any data as to whether or not it really makes a difference.”

Introduced in March 2009, Mercedes’ touts its Attention Assist technology as “accident prevention at its best,” but they haven’t provided any statistical evidence showing fewer sleep-related accidents in models equipped with the technology as compared to those without. The company’s internal research, however, showed that the technology, which consists of an audible alarm and a flashing coffee cup-shaped dashboard light, did have a positive effect.

In over 700 drivers, Attention Assist prompted “more than 50 percent of the drivers to take a break or change drivers, while another 10 percent woke up fellow passengers as a reaction,” according to Jochen Haab, a technical support manager at Mercedes-Benz headquarters in Germany.

Back in the early 2000s, Mercedes engineers contacted the sleep researcher David Dinges to help design a sleep detection system for their cars. Dinges, a scientist at the University of Pennsylvania School of Medicine, had been working with drowsiness detection technologies since the 1990s. His research showed that the speed of blinks was the best predictor of attention lapses — the more that drivers blinked in a performance test, the more likely they were to lose focus.

Despite their accuracy, eyelid-based systems are extremely difficult to design because of obstacles such as sunglasses, wandering-eyed drivers and variations in individual eye shape.

Dinges recently met with other sleep researchers at Stanford University to discuss the difficulties of making eyelid-based systems work well. Dr. William Dement, director of the Sleep Disorders Center at Stanford, agreed with Dinges that the secret is in the eyes: “Eyelids are the number one indicator [of drowsiness], but if you go beyond eyelids, the next best thing is steering.”

The change in steering behavior is the most important parameter in Mercedes’ system, according to the company’s Haab. In an email, he explained that Attention Assist uses a high-resolution steering wheel angle sensor accurate to one-tenth of a degree.

Mercedes’ technology works by creating an initial driver profile during the first 20 minutes of each trip, using this as a baseline against real-time sensors capable of assessing fatigue. If the system identifies a deviation consistent with drowsiness, the coffee cup flashes and the alarm sounds.

Drivers can temporarily clear the alarm, but if the system still detects drowsiness after 15 minutes it will go off again. Fifteen minutes might not seem very long, but Dement pointed out that “the moment you go from wake to sleep is almost instantaneous.”

Another weakness in Mercedes’ technology, according to some sleep researchers, is that Attention Assist doesn’t monitor driver performance below 50 mph. “Fall-asleep accidents can occur at any time, any place or any speed,” said Schwab.

Saab’s version, known as a Lane Detection Warning System, is similar to Mercedes’ — it detects driver inattention via turn signals, braking, accelerations and steering, according to John Libbos, product manager of Saab Cars North America.

Saab is working to perfect an infrared system to track eyelids, but “this is an extremely sophisticated technology,” Libbos noted. “There is still so much data that needs to be collected to make it as reliable as possible for real world situations.”

Sleep experts emphasize that in the real world, drivers will never be able to count on any alerting device to keep them awake in all circumstances. The safest way to drive drowsy is to not drive at all.

“It’s not a complicated issue, it’s just supremely important,” said Dement.  “It’s life and death. Sleep and die.”

About the Author

Allison T. McCann

Allison T. McCann has a B.S. in science, technology and society from Stanford University, an interdisciplinary major that sparked an interest in all things that swipe, zoom, beep, fly or otherwise involve cool technology. How does that work? is the eternal question that lingers in her thoughts during long runs along the Hudson. A born and raised California girl, she’s a bit nervous about the New York winters, but is very excited to continue writing about science and technology at SHERP. Follow her on Twitter.


1 Comment

Nick says:

I’d be interested to know what frequency they are running those infrared eye sensors. If perfected, it could easily be modified to assess drunk drivers as well. Great reporting!

Leave a Reply

Your email address will not be published. Required fields are marked *


The Scienceline Newsletter

Sign up for regular updates.