At the Sberbank’s Artificial Intelligence (AI) Laboratory the voice files are then converted into a spectrogram displaying sound energy across frequencies to be analyzed using a deep convolutional neural network (CNN).
A laboratory in Russia has designed an algorithm that detects diagnostic signals of COVID-19 within 60 seconds. This mainly depends on a symptom checker and three sound models which includes the voice, breathing and coughs. At the Sberbank’s Artificial Intelligence (AI) Laboratory the voice files are then converted into a spectrogram displaying sound energy across frequencies to be analyzed using a deep convolutional neural network (CNN). This CNN was trained on open data only, which included thousands of samples of breathing and coughs aggregated from COVID-positive patients in Russian clinics.
Last year in November Sberbank had announced their readiness to create an algorithm for detecting COVID.
In an official statement Alexander Vedyakhin, First Deputy Chairman of the Executive Board, Sberbank, has said, “Our model does not have the accuracy of the biological PCR-based research. But it already has similar characteristics. However, the model we have has adjustable sensitivity and is much easier to use. Also, it is not only convenient but cheaper too.”
He has also mentioned that it is not a medical diagnostic tool, but is good to be used as a daily personal checker as it takes just about 60 seconds to do a test and get a result.
According to Mr Vedyakhin, the Sberbank is getting ready to create a special application that will be available on App Store and Google Play. And in the process will further fine-tune the accuracy of the model.
Sberbank & AI
Last December it had hosted the largest Artificial Intelligence Journey ( AI Journey), which was an international conference on artificial intelligence and data science. It had attracted over 25,000 people from 87 countries to participate and a large number of top Indian companies had registered themselves for the event.