Scientists at La Trobe University have created a sophisticated artificial intelligence algorithm that can help understand if someone has drunk too much alcohol and is in a state of intoxication. We may soon see it in action instead of those booze-testing gadgets you find lying around.
All that is needed is for the system to listen to the voice of the "analysed" person: it only takes 12 seconds. The newly published studyI link it here) illustrates the implementation phases of this system, called ADLAIA. Abraham Albert Bonela, a doctoral student, conducted it under the supervision of professors Emmanuel Kuntsche e Zhenhe arriving to determine with incredible precision the state of drunkenness of a person through a few seconds of audio.
Because it is important
Have you ever heard of acute alcohol intoxication? Probably yes, but maybe you don't know how dangerous it is for your health (of those who are in a state of intoxication and also of others). Alcohol impairs our cognitive and psychomotor skills, increasing the risk of road accidents and acts of violence. Today, drunkenness is measured with an instrument called a breathalyzer, which measures a person's blood alcohol concentration. An algorithm that derives this data from the voice could produce a real breakthrough.
Thanks to this innovative technology it could be possible to monitor alcohol consumption more precisely and prevent accidents due to drunkenness: even lead to cars or voice assistants that tell us if we are drunk before getting behind the wheel, before entering a pub or a stadium, or maybe while we're still at home.
How the algorithm that measures drunkenness was developed
ADLAIA has been studied and tested on a dataset of 12.360 sound clips of people talking (both drunk and sober). At the moment, the algorithm is able to recognize people who have a blood alcohol level of at least 0,05% with an accuracy of 70%. If we raise the bar (with a blood alcohol content above 0,12%) the accuracy reaches 76%, but the system is constantly improving, and will soon reach decisive reliability.
Tell the truth: is it or isn't it a big step forward? Don't talk too much, though: I could change my mind about that beer I wanted to offer you.
