Sber's artificial intelligence laboratory has developed an algorithm for detecting coronavirus infection by sound. This was reported by RIA Novosti with reference to the company's press service. The algorithm uses a person's voice, breathing and coughing, and the user must also complete a short questionnaire survey. Sound files are transformed into a spectrogram, which shows the energy of the sound at different frequencies. The data is analyzed using a deep convolutional neural network. The examination takes about 60. The model does not achieve the accuracy of biological PCR, the screening developed is "not a medical diagnostic tool, but rather a personal daily checker."
      The app will soon be available on the App Store and Google Play. To train the algorithms, open data were used - more than a thousand samples of breath and cough sounds from patients diagnosed in Russian clinics. Earlier, researchers at the Massachusetts Institute of Technology announced the development of an algorithm to distinguish the cough of patients with coronavirus. In December 2020, the Sberbank group of companies and the Skolkovo Institute of Science and Technology announced a deal to create an ecosystem for the development of artificial intelligence in healthcare in Russia.