The coronavirus is a large family of viruses, but only six of them were known to affect humans. With this new virus, there are now seven. One of these causes Severe Acute Respiratory Syndrome (Sars) which in 2002 killed 774 people in China. Meanwhile, the new virus is what everyone already knows by now, the well-known Covid-19.
The whole world went through two very difficult years with the pandemic and research and vaccines being developed at the same time, but without much information about the virus that was new. Fortunately, the research paid off and vaccines arrived.
However, this does not mean that the world has become Covid-free. The virus still exists, it infects people, but thanks to vaccines, the death toll from it has dropped dramatically.
And as Covid is still a reality, a team of scientists, experts in artificial intelligence developed a mobile application that detects cases of Covid-19 in people’s voice. The most impressive thing is that this app has greater accuracy and is faster than the antigens used so far.
This system was presented on Monday of this week at the International Congress of the European Respiratory Society in Barcelona, Spain. In addition to accuracy, the app is cheaper than antigen tests. Because of this, it can be used in low-income countries, where tests are expensive or difficult to perform.
According to Wafaa Aljbawi, a researcher at the Institute of Data Science at the University of Maastricht (Netherlands), the artificial intelligence model has an accuracy of 89%, which varies by brand.
“Our results are promising and suggest that voice recordings and tuned AI algorithms can be highly accurate in determining which patients have Covid-19. These tests are free and easy to interpret. In addition, they can be remote virtual tests and their response time is less than a minute so they can be used, for example, at entry points of large agglomerations to ensure rapid detection,” he said.
This identification is possible because Covid normally affects the upper respiratory tract and vocal cords. This causes a change in the voice of the infected person.
Knowing this, Aljbawi, Sami Simons, a pulmonologist at Maastricht University Medical Center, and Visara Urovi, from the Data Science Institute, went to study to see if it was possible to use artificial intelligence to analyze people’s voices and detect Covid infections.
For this, they used the Covid-19 Sounds open app, created by the University of Cambridge to study the symptoms of the coronavirus. It consists of 893 audio samples from 4,352 healthy and unhealthy participants. Of those, 308 tested positive for Covid-19.
The application is installed on the person’s cell phone and then they must provide basic information, data about their medical history and habits, such as smoking. After that, the app asks her to record her breathing, her cough, and her voice.
Then, through the voice analysis technique called Mel’s spectrogram analysis, which can identify different characteristics of the voice, such as volume, power and variation, the app is able to decompose the properties of people’s voices.
To differentiate the voice of who has Covid and who doesn’t, scientists built different models of artificial intelligence and studied which one worked best in this differentiation.
The model that reached 89% accuracy was the Long-Short Term Memory (LSTM), based on neural networks that mimic the way the human brain operates and is able to recognize the underlying relationships in data.
These results still need to be validated in a larger survey, which will be done with 53,449 audio samples from 36,116 participants.
Images: UOL, One like
#App #detects #Covid #based #peoples #voice #accuracy