How a new technology that detects aberrant alcohol-related behavior could save lives
Researchers at La Trobe University have developed an artificial intelligence (AI) algorithm that can be used in conjunction with expensive and potentially biased breath test equipment in pubs and clubs.
With just a 12-second voice recording, the technology can instantly determine if you’re over the legal alcohol limit.
In a paper published in the journal Alcohol, the research was led by a Ph.D. Student Abraham Albert Bonera is supervised by Prof. Emanuel Kuntsch and Associate Prof. Zheng He of the Center for Research on Alcohol Policy and the Faculty of Computer Science and Information Technology at La Trobe University, where he develops audio-based deep learning algorithms. I’m talking about development. Identifying Alcohol Intoxication (ADLAIA) allows you to determine an individual’s intoxication status based on a 12-second speech recording.
According to Albert Bonella, acute alcohol intoxication impairs cognitive and psychomotor skills and leads to a variety of public health hazards, including road accidents and alcohol-related violence.
“Intoxicated individuals are typically identified by measuring their blood alcohol concentration (BAC) using expensive and labor-intensive breathalyzers,” said Albert Bonela.
“A test that could rely on someone just speaking into a microphone would be a game changer.”
The algorithm was developed and tested using a database dataset of 12,360 audio clips of intoxicated and drunken speakers. According to the researchers, ADLAIA was able to identify intoxicated speakers with a BAC of 0.05% or higher with nearly 70% accuracy. The algorithm showed a high performance of almost 76% in identifying intoxicated speakers with a BAC greater than 0.12%.
Researchers believe that one potential future application of ADLAIA is its integration into mobile applications, where it can be used in environments (such as bars and sports stadiums) to obtain immediate results on an individual’s drunkenness. suggesting.
Albert Bonera said, “If we could identify intoxicated individuals based on speech alone, breath-based alcohol testing would be a much cheaper alternative to current systems, which are often costly and unreliable. It will be a great alternative,” he said.
“If the overall performance improves further, ADLAIA could be integrated into mobile applications and used as a preliminary tool for identifying alcoholic individuals.”
Original: Deep learning algorithms can hear alcohol in your voice
Than: La Trobe University
