A smartphone app that can detect COVID-19 in voices using AI

The method used is called Melspectogram analysis which identifies different voice features

At the European Respiratory Society International Congress in Barcelona, Spain, researchers introduced a smartphone app that can detect COVID-19 infection through people's voices with the help of Artificial Intelligence (AI). They claim the app is significantly accurate than most antigen tests and is also an easy, quick, and cheaper method than tests. This means it can benefit lower-income countries where PCR tests are expensive and challenging to distribute. Wafaa Aljbawi, a researcher at the Institute of Data Science, Maastricht University, The Netherlands, said that “the promising results suggest that simple voice recordings and fine-tuned AI algorithms can achieve high precision in determining which patients have COVID-19 infection."

COVID-19 usually affects the vocal cords and the upper respiratory tract, causing changes to a person's voice. The researchers decided to investigate if the difference could be identified using AI to analyze the voices and detect COVID-19. The research was based on the data from the University of Cambridge's crowdsourcing COVID-19 Sounds App. The application contains 893 audio samples from healthy and unhealthy participants, 308 of whom had tested positive. The method used is called Melspectogram analysis which identifies different voice features such as power, variation over time, and power.