A study carried out at Stanford University shows that it is possible determine if a person is drunk or ascertain, with a degree of reliability equal to 98%, whether his/her blood alcohol level is higher than the legal limits. The results of the experiment, which involved both young 18-year-olds and individuals aged 21 and over, surprised even the promoters of the initiative.
Leveraging the progress made in the field ofsignal processingin acoustic analysis and inmachine learningthe researchers were able to develop a model capable of detecting voice alterations in all subjects whose blood alcohol level is equal to or higher than that permitted for driving a vehicle.
Determine if a person is drunk: it can be done with a simple smartphone
Professor Brian Suffoletto, associate professor of medicine at Stanford explains that the idea was to develop an instrument capable of passively sample data of an individual while carrying out his routine daily and probe for changes that could indicate an episode of alcohol consumption.
Given that further investigations are necessary to confirm the validity of the results, the 98% accuracy it is a sensational result. The academic work carried out so far has the potential to offer a “intervento just-in-time”, useful to prevent road accidents and deaths related to excessive alcohol intake.
L’analysis of vocal characteristics and alterations in speech that can be carried out with a simple smartphone, are also able to detect situations border line, or cases in which the intake of alcoholic beverages has caused an increase in the blood alcohol level around the legal limit value. In these situations the state of inebriation could become manifest only with specific tests: i.e. with breath analysis or blood tests. Yet, as scholars say, the use of vocal models and the machine learning are able to highlight potential alert situations.
The mechanism developed by Suffoletto and his collaborators proved to be extremely precise, also thanks to the use of a series of technical precautions: the isolation of the voice and the carrying out of measurements on frequency and tone, with sampling carried out every second.
The smartphone becomes a sensor based on artificial intelligence
When a mechanism based onartificial intelligence enters the private sphere of the individual, it is obviously necessary to tread with the proverbial lead feet. In fact, it is unthinkable to develop a system that is always listening and provides information in a completely automated manner.
What might be possible to achieve, and this is what we are talking about at Stanford, is an intelligent system capable of carrying out a real-time verification on the state of an individual on demand, or at the express request of the interested party. Or, you could make sure that depending on where the user is locatedfor example a bar, a pub, a nightclub, a notification will appear inviting you to voluntarily take a test.
In the case of minors, however, a parent could possibly request the carrying out of remote testing. The applications of parental control they already offer the possibility of limiting the user’s “range of action” regarding the visit of websites, the installation and use of apps.
In Europe, with the presentation of the bill known as Chat Control 2.0, there was a real risk of “mass surveillance”. It goes without saying, therefore, that technology must be at the service of everyone, in full respect of privacy and individual rights. Certainly not an imposition.