First Deputy Chairman of the Board of Sberbank, Alexander Vedyakhin, in an interview with RT, spoke about the work of an algorithm that can determine the presence of coronavirus by coughing.
According to him, open data was used to train the program - more than a thousand samples of breathing and coughing sounds collected with the consent of patients.
“We have launched a very complex large neural network, which distinguishes by the semitones of the sound spectrogram what is indistinguishable to the human ear. She distinguishes how these spectrograms distinguish a person who is sick with COVID-19 from a healthy person or someone who coughs because of a simple cold or flu,”Vedyakhin said.
He stressed that the application allows you to predict with a high degree of probability whether a person has a coronavirus, but it is not medical and does not cancel the delivery of a PCR test.
Vedyakhin also told how the application will work. So, after switching on, it will be necessary to answer several questions related to the presence of temperature and contacts with a sick COVID-19, then cough for a few seconds, actively breathe and talk, after which the program will issue a conclusion about the likelihood of infection.
Earlier, Sber said that he had developed an algorithm for detecting coronavirus in a minute by coughing.