A team of researchers from Stanford Medicine and the University of Toronto has determined that sensors in smartphones and smart speakers could potentially reveal how intoxicated a person is based on changes in the person’s voice.

To make this determination, the researchers conducted a study of adults who were issued a weight-based amount of alcohol and were subsequently assigned tongue twisters both one hour before they consumed the alcohol and every 30 minutes for up to seven hours after they consumed alcohol.

Source: Public domain

As the participants read their assigned tongue twisters aloud, a smartphone placed within 1 ft to 2 ft away from the subject recorded their voice. Meanwhile, the subject’s breath alcohol concentration was measured at each hourly interval.

Further, the researchers employed digital programs capable of isolating the speaker’s voice and breaking it up into one second increments, which were subsequently analyzed for indicators of changing voice patterns such as pitch and frequency.

When measured against breath alcohol concentrations, the researchers reported that ongoing changes in a speaker’s voice patterns during the study predicted intoxication with almost 100% accuracy.

The researchers suggest that such a tool, combined with other sensors for measuring gait, voice and texting behavior, for instance, might lead to a device that could potentially prevent intoxicated driving in the future.

An article detailing the team’s findings, “E. Detection of alcohol intoxication using voice features: A controlled laboratory study,” was detailed in the Journal of Studies on Alcohol and Drugs.