Cars could detect drunk and impaired drivers using technology developed by U-M engineers

The most recent data available from the National Highway Traffic Safety Association (NHTSA) shows that drunk driving remains one of the leading causes of death on U.S. roads. NHTSA reports that 13,384 deaths were linked to drunk driving in 2021, and an average of 37 people die each day in drunk driving accidents. In an effort to lower those statistics, University of Michigan (U-M) engineers have developed new technology that could recognize impaired drivers.

"What we're trying to do is use camera-based approaches that you probably carry in your pocket or pocketbook to look at the state of the driver to see if they are impaired," says Mohammed Islam, a U-M professor of electrical engineering and computer science. "We can use that same kind of technology in the car to see if we can get a bit closer to recognizing the state of a driver."

Islam, who is the leader of U-M's impaired-driver detection project, explains that many newer vehicles already have Advanced Driver Assistance Systems (ADAS) cameras that can monitor a driver's alertness. Islam’s team proposes augmenting existing ADAS cameras with infrared Light Detection and Ranging (LiDAR) or structured light 3D cameras costing a modest $5-$10.  

"Market studies show that more than 40 million camera-based systems will be deployed in automobiles by 2027, so it's a direction that's already happening," he says. "We're just doing incremental cost increase to adapt that system and use it for things like impairments."

The U-M team's proof of concept experiments, which interpret data captured by LiDAR cameras with artificial intelligence tools, can detect five markers that could indicate that a driver should not be behind the wheel. After taking a 20- to 30-second video of the driver's face, the system assesses blood flow, increased heart rate, eye behavior, body posture and head position, and respiratory rate. 

"We're looking for telltale signs that you probably intuitively recognize as correlated to being drunk," Islam says. "We're trying to do the same process, but in our case, we're using artificial intelligence machine learning to help the process."

He explains that the system does what's called "anomaly occurrence detection," similar to the way that a credit card company might detect fraudulent transactions. For instance, a financial institution might have baseline data on normal spending for a customer. Should the same customer make a sudden number of large charges, perhaps in a place that they've never gone before, then fraud detection alerts would be triggered. 

Islam is seeking more local interest and financial support to get the technology out of the university and onto the state's roads. It's "a win-win," he says, because his team needs to grow the technology, and potential partners could have the competitive advantage of cars that have more safety features for their buyers. 

"We sit in the U.S. car capital. This is where technology innovations for the car companies happen. Henry Ford, General Motors, and Chrysler are all in this area," he says. "So, by golly gosh, it should be the auto capital that helps to at least contribute to remedying this problem."

Jaishree Drepaul is a freelance writer and editor based in Ann Arbor. She can be reached at

Photo by Jeremy Little, Michigan Engineering.
Enjoy this story? Sign up for free solutions-based reporting in your inbox each week.