Could Your Voice Show Signs of COVID Infection?

New research points to speech analysis for identifying asymptomatic carriers.

Posted Jul 15, 2020

By now it’s old news that speaking increases the possibility of airborne transmission of COVID 19 through aerosolized droplets. But what if the sound of your voice could be used to identify early COVID infection? New research from MIT scientists suggests it might be another clue, like the loss of taste and smell, to asymptomatic carriers or a sign of early infection.

Taking advantage of audio data from celebrities who unknowingly had COVID infections, scientists associated with MIT’s Lincoln Laboratories report that they were able to track changes to voice quality as a result of infection. The big advantage of locating this type of biomarker is that acoustic assessment of voice changes could provide a non-invasive way, like taking a temperature or measuring oxygen levels, to identify someone early in the stages (or asymptomatic with) viral infection.

In a testament to the power of social media transcending its usual purview and our new reliance on virtual research subjects in the days of COVID, the researchers used recordings of five celebrities from sites such as Twitter, Instagram, and YouTube to compare their pre-COVID voice with their speech when they had contracted COVID but were asymptomatic.

Their hypothesis that our voice might be affected by infection is based on the interdependence between respiratory and speech systems. What we use for speaking — the lungs, trachea, larynx (or voice box) and the mouth and nose — are also used for breathing, which is why we sound raspy or ‘stuffed-up’ when we have a cold or flu.  

Ian Clayton, used with permission
Pitch and articulation tracking for the word 'COVID'
Source: Ian Clayton, used with permission

To speak, lung air is pushed out and shaped by our vocal apparatus into specific sounds as it moves toward our mouths in a coordinated process. Though simplifying somewhat, we can measure the acoustic signatures of distinct vowels, consonants, pitch, and other related aspects in the soundwaves of our emitted speech and, in turn, make predictions about the shape and state of the vocal tract that produced it.  

The MIT lab’s research is predicated on the concept that acoustic changes will go hand in hand with illness-related changes in the size, shape, or mechanics of your respiratory and articulatory system. With COVID-induced inflammation of the respiratory system and its deleterious effects on motor coordination in the vocal tract, the researchers indeed found that infection altered the acoustics (the soundwaves) of the celebrity’s voice when he or she was infected. 

The big what-if, of course, is whether such a finding can be used in an accessible way to track infections in populations like schoolchildren or sports fans coming to watch a big game (or any game). At this point, it is too early to tell whether this work will translate into an effective screening tool, as a number of other influences on voice quality (like hormonal changes, speaking style, emotional states) might be confounding the results.

But if further research validates their findings, the acoustic software needed to acquire measurements is widely available and, the researchers suggest, could potentially be applied via a smartphone app similar to an existing one developed to help identify vocal biomarkers of depression. The COVID-19 based app could compare a baseline speech sample to new samples at regular intervals, offering a quick and simple way to check for early or unknown infection.  Given the ubiquity of cellphones and our vast experience talking into them, this new app might just be as popular as the social media outlets that helped inspire it.

References

T. Quatieri, T. Talkar and J. Palmer, "A Framework for Biomarkers of COVID-19 Based on Coordination of Speech-Production Subsystems," in IEEE Open Journal of Engineering in Medicine and Biology