New Deep Learning Algorithm Can Tell if You’re Above the Legal Drinking Limit
AI and Data Science Newsposted by ODSC Team January 12, 2023 ODSC Team
A new deep learning algorithm just needs 12 seconds to determine if you’re above the legal drinking limit. This comes to us from a paper published in Science Direct which states that La Trobe University researchers developed an algorithm that only needs about 12 seconds of audio to make a determination on blood alcohol count levels. The purpose of the program is to complement other more expensive breath-testing technologies.
The study’s leader is Ph.D. student Abraham Albert Bonela, under the supervision of Professors Emmanuel Kuntsche from the Center of Alcohol Policy Research and Associate Professor Zhen He from Research and the Department of Computer Science and Information Technology at La Trobe University. The audio-based deep learning algorithm, ADLAIA, was trained to detect and identify alcohol inebriation levels based on a 12-second clip of their speech.
How this algorithm works is interesting. First, it was developed and trained against a database dataset of over 12,000 audio clips from both sober and inebriated speakers. So far the deep learning algorithm is able to identify inebriated speakers with an accuracy of almost 70%. The program performed better the higher the BAC count of the speaker was, moving closer to 80%. Though it’s still early in the process, the algorithm’s results are quite promising. And with that promise comes very interesting possibilities according to Albert Bonela, “A test that could simply rely on someone speaking into a microphone would be a game changer.”
Imagine a simple, low-cost, and accurate program, that can determine with high accuracy if someone was above the legal limit. It could be an impressive tool when it comes to public safety. Looking to the future, the study’s lead is also looking into the possibility of algorithms integration into mobile apps. “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.” This could provide not only law enforcement with another tool, but private establishments such as bars, pubs, etc. with a means to easily test if patrons are over the legal limit before ending service or even calling for a pick up so they don’t drink and drive.
The exposition of AI technology in medicine seems to be continuing early into 2023. It will be interesting to see what AI and deep learning technologies will bring us this year.