Researchers at Stanford Medicine and the University of Toronto have created a process that allows smartphones' sensors and speakers to determine how drunk a user is, a technology that could prevent intoxicated people from driving under the influence of alcohol.
The experiments revealed that changes in subjects' vocal patterns could be used by the devices to accurately predict alcohol intoxication 98% of the time.
This is according to a study carried out with 18 participants aged 21 or over. In addition to being given a weight-based dosage of alcohol, participants were asked to randomly perform a series of tongue twisters: one before they drink and one every hour after consuming the alcohol over a seven-hour period.
After reading the tongue twister aloud, participants were instructed to record their voices on a smartphone that was placed on a table located one and a half meters away. The alcohol content in the breath was also documented by the researchers at the beginning of the experiment and then every 30 minutes after the first drink, for a maximum of seven hours.
The accuracy of our model really caught me by surprise.
While we are not pioneers in highlighting changes in speech characteristics during alcohol intoxication, I firmly believe that our superior accuracy stems from the application of cutting-edge advances in signal processing, acoustic analysis, and machine learning.
Said principal investigator Brian Suffoletto, physician and associate professor of emergency medicine at Stanford.
According to Suffoletto, the goal of this type of study is to provide interventions “just-in-time” to prevent injuries and deaths caused by road accidents or other accidents related to alcohol poisoning. He explained that the ideal intervention tool for these types of incidents would be simple to use and easily accessible.
Given the widespread use of smartphones and smart speakers, these devices are an excellent option for warning individuals when they are becoming too intoxicated and should not drive or may even require hospitalization.
Imagine if we had a tool capable of passively collecting data from an individual during their daily routines and looking for changes that could indicate an episode of alcohol consumption to know when they need help.
Suffoletto said.
The researchers' ultimate goal is to create an intervention system that can help save lives and prevent injuries, making it so unobtrusive that people are actually willing to use it.
O study was published in the Journal of Studies on Alcohol and Drugs.
