A new AI technology can instantly determine whether a person is above the legal blood alcohol limit. The AI proceeds by analyzing a 12-second clip of the voice.
La Trobe University researchers develop artificial intelligence algorithm (AI) algorithm that could work in parallel with alcohol testing devices. in pubs and clubs.
AI algorithm is based on voice recognition
In a publication in the journal Alcohol, doctoral student Abraham Albert Bonela describes the development of theaudio-based deep learning algorithm to identify drunkenness. Dubbed ADLAIA, this algorithm led to the design of an AI capable of determining drunkenness from a 12-second recording of a person’s speech.
Professor Emmanuel Kuntsche supervised Albert Bonela in the validation of the study. Kuntche works at the Center for Alcohol Policy Research.
In addition to detecting drunkenness, the technology could also reveal speech and articulation defects. This would make it possible to identify people at risk of developing communication-related health problems.
AI voice-activated intoxication detection: an algorithm as good as an alcoholmeter
The algorithm was tested on a dataset containing about 30,000 records of patients seeking treatment for alcohol abuse.
The results showed thatADLAIA had excellent accuracy in identifying intoxication – 90% or more – compared with other methods used to detect intoxication.
According to the researchers, the AI was able to identify drunken speakers – with a blood alcohol content of 0.05% or more – with a accuracy of almost 70%.. The algorithm achieved a superior performance of almost 76% in the identification of intoxicated speakers.
According to Albert Bonel, today’s alcohol detection tools are expensive and labor-intensive. A test that simply relies on someone speaking into a microphone would be a game-changer.
AI alcoholmeter available on mobile application
The researchers suggest that a potential application for ADLAIA would be in public spaces. The AI would then provide instantaneous results on an individual’s state of inebriation. According to Albert Bonela, ADLAIA could be integrated into a mobile application for accessibility to the general public.
This would avoid situations where people get drunk and end up causing problems. The researcher looks forward to seeing how this technology develops and what applications it might have in the future.
NowadaysAI contributes more and more to the world around us. Programs are learning to do everything, from reading to writing to driving cars.