acm-header
Sign In

Communications of the ACM

ACM TechNews

Deep Learning Algorithm Can Hear Alcohol in Voice


View as: Print Mobile App Share:
A man having a drink with friends.

La Trobe Universit's Abraham Albert Bonela observed that intoxicated individuals are usually identified by measuring their blood alcohol concentration; "A test that could simply rely on someone speaking into a microphone would be a game changer."

Credit: La Trobe University

Researchers at Australia's La Trobe University have developed an algorithm that can instantly determine whether a person has exceeded the legal alcohol limit based on a 12-second recording of that person's voice.

The Audio-based Deep Learning Algorithm to Identify Alcohol Inebriation (ADLAIA) was developed with, and tested against, a dataset of 12,360 audio clips of inebriated and sober speakers.

ADLAIA was able to identify inebriated speakers having a blood alcohol content (BAC) of 0.05% or higher with an accuracy of nearly 70%, which climbed to 76% for speakers with a BAC of more than 0.12%.

Said La Trobe's Abraham Albert Bonela, "Upon further improvement in its overall performance, ADLAIA could be integrated into mobile applications and used as a preliminary tool for identifying alcohol-inebriated individuals."

From La Trobe University (Australia)
View Full Article

 

Abstracts Copyright © 2023 SmithBucklin, Washington, DC, USA


 

No entries found