COVID-19 signs may be hidden in the voice

Compartir : Facebook Twitter Whatsapp

According to an experiment by researchers at the Massachusetts Institute of Technology, it is possible to search for vocal biomarkers for Covid-19 by analyzing the sounds made by asymptomatic people when they speak.

The infection would have the ability to affect the proper functioning of the muscles necessary for speech, even in asymptomatic people.

It's often easy to tell when colleagues have a severe cold - their stuffy nose gives them a lower voice or even a nasal tone. Infections can affect the quality of our voices in several ways, but is this the case with the new coronavirus? It is now known that the infection it causes can cause loss of taste (ageusia) and / or smell (anosmia), but researchers at the Lincoln Laboratory of the Massachusetts Institute of Technology (MIT) wanted to know in a study whether they can be detected changes in the voices of these patients, even when these changes are too subtle to be heard or noticed by the people around them.

By processing the voice recordings of people infected with COVID-19 but not yet showing symptoms, these researchers found evidence of vocal biomarkers, or measurable indicators, of the disease. These biomarkers come from disorders caused by infection in the muscles of the respiratory, laryngeal and articulator systems. Although this research is still in its infancy, the first results establish a framework for studying these vocal changes in more detail. Furthermore, this discovery may be promising for the use of mobile applications to detect affected people, particularly those who are asymptomatic.

Inflammation of the respiratory system changes the voice.

"When symptoms appear, a person generally has difficulty breathing. Inflammation of the respiratory system affects the intensity with which air is exhaled when speaking, "the researchers explain." This air interacts with other potentially inflamed muscles on their journey toward speech production. "These interactions have an impact on volume , tone, stability and resonance of the voice ". To carry out their study, the scientists searched YouTube for videos of five celebrities or television presenters who gave interviews when they were COVID-19 positive but asymptomatic.

They then searched and downloaded interviews with these people before the COVID-19 epidemic hit, with the best possible audio conditions, and then used algorithms to extract characteristics of the voice signals from each audio sample. "These vocal characteristics serve as agents for the movements underlying speech production systems," said Professor Tanya Talkar, who participated in the study. Comparing the variety of sounds produced in the two types of audio excerpts (with or without infection), they noted less complexity of voice quality in the COVID-19 interviews compared to previous interviews.

A new type of detection with multiple possibilities.

"These preliminary results suggest that biomarkers derived from coordination of the vocal system may indicate the presence of COVID-19," say the researchers, for whom the next step will be to work on real audio samples from people with positive results for COVID-19. Beyond collecting additional data to fuel this research, the team plans to use mobile apps to implement it.

In particular, they want to integrate COVID-19 voice detection into the VoiceUp app, initially developed by the McGovern Institute for Brain Research to study the link between voice and depression.

"An in-app detection system could detect infections early, before people get sick or from people without symptoms," said Professor Jeffrey Palmer, who led the research group. "Even after a diagnosis, this detection capability could help clinicians remotely monitor their patients' progress or the effects of a vaccine or drug treatment. The main difficulty to overcome is discovering how to deal with confounding factors, those that cause vocal changes such as the recording environment, the patient's emotional state or other illnesses.

Compartir : Facebook Twitter Whatsapp