When someone has a cold, the person speaks through their nose, is hoarse: their illness can be heard in their voice. The algorithm works on the same principle but is much more precise, explains Mégane Pizzimenti, scientific manager of the Colive Voice study, conducted by the Luxembourg Health Institute. “With the help of ‘machine learning’ algorithms, we are able to extract more than 6,000 voice characteristics from a voice recording. It is thanks to them that we will be able to see at a given moment given if we have something specific to a person who has a disease or a symptom”, she explains. A study launched in 2020 by a group of international researchers.
Participants from around the world called to send voice recordings
The algorithm is only in its testing phase. To train it and keep it learning, scientists want to enrich the database. From a smartphone, everyone, sick or not, can participate by sending voice recordings to the study site, six different languages are possible. “We will add new languages as Colive Voice grows because we are looking for participants from all over the world with different accents to develop the most precise biomarkers possible”, continues Mégane Pizzimenti. Everyone can therefore participate by going to www.Colivevoice.org .
For the moment the first results are conclusive, particularly concerning diabetes. The algorithm even detects whether it is type 1 or type 2 diabetes.
To view this Youtube content, you must accept cookies Advertisement.
These cookies allow our partners to offer you personalized advertising and content based on your browsing, your profile and your interests.
Manage my choices I authorize