Experts of the financial institution’s artificial intelligence laboratory note that their model has not yet reached the accuracy of the PCR test.
The epidemic of coronavirus
Employees of the Sber Artificial Intelligence Laboratory have developed an algorithm to detect the coronavirus in one minute. It is noted that the neural network is able to determine the probability of infection with COVID-19 by sound. The presence of a new extracellular agent in a person is determined by the results of a short survey on the symptoms of the disease and three sound models — voice, breathing and cough.
Experts explained that audio files are converted into a spectrogram, which clearly displays energy at different frequencies. After that, a deep convolutional neural network analyzes the sound samples. The developers noted that their model has not yet achieved the accuracy of a biological PCR sample.
“It makes it possible to make a configurable sensitivity, much easier to handle, more convenient and, importantly, cheaper. This is not a medical diagnostic tool, but rather a personal daily checker — passing the test and getting the result take only 60 seconds, ” RIA Novosti quotes the first deputy chairman of the Board of Sberbank Alexander Vedyakhin.
In the foreseeable future, Sberbank will create a corresponding application for users, which will be available for download in the App Store and Google Play