Researchers at La Trobe University have created an artificial intelligence (AI) algorithm that might be used at pubs and clubs in addition to pricey and potentially biased breathalyzers.
An audio recording of a person’s speech for 12 seconds can be used by the technology to instantaneously assess whether they have consumed more alcohol than is permitted.
The development of the Audio-based Deep Learning Algorithm to Identify Alcohol Inebriation (ADLAIA), which can assess a person’s level of intoxication based on a 12-second recording, is described in the study led by Ph.D. candidate Abraham Albert Bonela and supervised by Professors Emmanuel Kuntsche and Associate Professor Zhen He from the Center for Alcohol Policy Research and the Department of Computer Science and Information Technology at La Trobe University, respectively.
Albert Bonela claims that acute alcohol intoxication impairs cognitive and psychomotor functions, which can result in a number of public health risks, including automobile accidents and alcohol-related violence.
According to Albert Bonela, intoxicated persons are typically recognized by testing their blood alcohol concentration (BAC) using expensive and labor-intensive breathalyzers.
A test that could only be conducted by having a subject talk into a microphone would be revolutionary.
Using a database collection of 12,360 audio samples of drunk and sober speakers, the algorithm was created and evaluated. The researchers found that ADLAIA has an accuracy of over 70% for detecting intoxicated speakers with BACs of 0.05% or above. With a BAC of more than 0.12%, the system performed better than 76% of the time in identifying intoxicated speakers.
According to the researchers, one potential use for ADLAIA in the future could be its integration into mobile applications and use in settings (such as pubs and sporting venues) to obtain fast results about people’s levels of alcohol consumption.
The existing procedures, where breath-based alcohol testing in these locations is expensive and frequently unreliable, would be significantly more affordable if inebriated people could be identified merely based on their speech, according to Albert Bonela.
ADLAIA “may be integrated into mobile applications and employed as a preliminary tool for recognizing alcohol-inebriated persons” upon “further improvement in its overall effectiveness.
