Researchers from the Lincoln Laboratory have found early evidence that subtle changes in peoples' voices may indicate whether they have been infected by COVID-19- even before they develop symptoms.
The idea that there may be vocal biomarkers for COVID-19 came to Thomas Quartieri, a senior staff member at the lab while watching the news. He and his team have previously found vocal biomarkers in people suffering from neurological disorders like ALS and Parkinson's. Given how COVID-19 affects the airways, and thus potentially speech production, they set out to investigate its effects.
To do so, he and his team searched YouTube for clips of celebrities and TV hosts who had given interviews while infected with the virus and asymptomatic. They also sought out videos to compare their voices before infection. In total, they gathered data on five subjects.
To detect speech differences, they used algorithms to extract features from the voice signals of each participant's audio. Pitch and steadiness of pitch were used as a proxy for the stability of vocal cords. Meanwhile, articulator movements (movement of the tongue, lips, jaw, etc.) were deduced from vocal quality, and movement in the respiratory system was measured by the amplitude of each signal or 'how loud' it was.
In the end, they found that subjects tended to have less movement complexity in their speech after contracting COVID-19.
"The coupling was less prominent between larynx and articulator motion, but we're seeing a reduction in complexity between respiratory and larynx motion," says Tanya Talkar, one of the study's authors.
Although this early research suggests a link between the vocal system and COVID-19 infection, the researchers say that more data is needed before conclusions can be drawn. To this end, they are now working with a publicly released dataset from Carnegie Mellon University containing more audio samples of people with COVID-19.
"A sensing system integrated into a mobile app could pick up on infections early before people feel sick or, especially, for these subsets of people who don't ever feel sick or show symptoms. This is also something the U.S. Army is interested in as part of a holistic COVID-19 monitoring system." says Jeffrey Palmer, lead author of the study.