Ben-Gurion University of the Negev01.06.17
A new “virtual breathalyzer” developed by a BGU researcher uses sensors in smartphones, smartwatches, fitness bands and virtual glasses to measure changes in gait that indicate intoxication levels with identical accuracy as police breathalyzer tests.
According to the U.S. Center for Disease Control, in 2013, one person died every 51 minutes in a motor vehicle accident caused by an alcohol-impaired driver.
"Alcohol distinctly affects movement, gait and balance in ways that can be detected by the built-in motion sensors on devices people carry around with them all the time," says Ben Nassi, a Master of Science student at BGU's Department of Software and Information Systems Engineering, who developed the device. “Our system simply takes a baseline reading while walking from the car to the bar and another one on the way back to compare and identify movements that indicate drunkenness.”
Applications based on Nassi’s trained machine learning model for measuring intoxication could be used to alert people, or even a connected car, and prevent users from driving under the influence.
In the study, Nassi and his team collected test data from patrons at different bars on five nights. They asked 30 participants (60 percent men, 40 percent women) to measure their gait before drinking and then 15 minutes after their last drink, which is the same standard used for police breathalyzers. Most of the study participants were in their early twenties, which is the group considered by the U.S. National Highway Traffic Safety Administration to have the highest risk of causing fatal accidents due to alcohol consumption.
Participants wore Google Glass augmented reality glasses, an LG G-watch on their left hand, a Microsoft Band on their right hand, and carried a Samsung Galaxy S4 cell phone in their right rear pocket. Each person walked for 16 seconds until they heard a beep through their headphones. Test results validated with a police breathalyzer detected intoxication levels with 100 percent accuracy.
"While the experiment used all four devices to measure movements in different parts of the body, a combination of watch and smartphone readings taken from at least two parts of the body yields similar results," Nassi said.
Smart wearable devices are a burgeoning market, with 275 million sold in 2016, and another 322 million units forecast in 2017. The researchers are optimistic that within a few years, the application will be useful for people who routinely use a smartwatch along with their smartphone.
“A system based on our approach could prevent a person from driving under the influence after an alert unobtrusively detects intoxication while they are walking to their car,” said Nassi. “As the Internet of Things (IoT) progresses, the system could even trigger a connected car not to start when a driver tests above the legal limit.”
Nassi worked with his advisors, Professors Yuval Elovici and Lior Rokach of BGU’s Department of Software and Information Systems Engineering on his Virtual Breathalyzer project, which has been uploaded to Arxiv.
According to the U.S. Center for Disease Control, in 2013, one person died every 51 minutes in a motor vehicle accident caused by an alcohol-impaired driver.
"Alcohol distinctly affects movement, gait and balance in ways that can be detected by the built-in motion sensors on devices people carry around with them all the time," says Ben Nassi, a Master of Science student at BGU's Department of Software and Information Systems Engineering, who developed the device. “Our system simply takes a baseline reading while walking from the car to the bar and another one on the way back to compare and identify movements that indicate drunkenness.”
Applications based on Nassi’s trained machine learning model for measuring intoxication could be used to alert people, or even a connected car, and prevent users from driving under the influence.
In the study, Nassi and his team collected test data from patrons at different bars on five nights. They asked 30 participants (60 percent men, 40 percent women) to measure their gait before drinking and then 15 minutes after their last drink, which is the same standard used for police breathalyzers. Most of the study participants were in their early twenties, which is the group considered by the U.S. National Highway Traffic Safety Administration to have the highest risk of causing fatal accidents due to alcohol consumption.
Participants wore Google Glass augmented reality glasses, an LG G-watch on their left hand, a Microsoft Band on their right hand, and carried a Samsung Galaxy S4 cell phone in their right rear pocket. Each person walked for 16 seconds until they heard a beep through their headphones. Test results validated with a police breathalyzer detected intoxication levels with 100 percent accuracy.
"While the experiment used all four devices to measure movements in different parts of the body, a combination of watch and smartphone readings taken from at least two parts of the body yields similar results," Nassi said.
Smart wearable devices are a burgeoning market, with 275 million sold in 2016, and another 322 million units forecast in 2017. The researchers are optimistic that within a few years, the application will be useful for people who routinely use a smartwatch along with their smartphone.
“A system based on our approach could prevent a person from driving under the influence after an alert unobtrusively detects intoxication while they are walking to their car,” said Nassi. “As the Internet of Things (IoT) progresses, the system could even trigger a connected car not to start when a driver tests above the legal limit.”
Nassi worked with his advisors, Professors Yuval Elovici and Lior Rokach of BGU’s Department of Software and Information Systems Engineering on his Virtual Breathalyzer project, which has been uploaded to Arxiv.
