This article is more than 1 year old

Thankfully, our AI savior is here to nail the COVID-19 pandemic: A neural network that can detect coughing

False positive rates, we've heard of them

The AI community is attempting to tackle the coronavirus pandemic using all sorts of algorithms in its toolbox. As such, it's coming up with ways to predict the virus’ protein structure, crunch through thousands of science papers, and now, er, detecting coughs.

Humans are pretty good at hearing coughs, so why do we need machine learning? Well, computers can automate the process of counting the number of nearby coughs, ultimately helping experts to gather statistics, and forecast and combat the spread of respiratory illnesses – whether it’s the common cold, festive flu, or even possibly COVID-19 or SARS – according to researchers at the University of Massachusetts Amherst (UMass) in the US.

For example, take FluSense: something the uni team described as a “contactless syndromic surveillance platform” in a paper published in the Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies this month. Essentially, it’s a device that counts the number of people coughing nearby.

And here's the university's PR blurb for the tech:

The FluSense creators say the new edge-computing platform, envisioned for use in hospitals, healthcare waiting rooms and larger public spaces, may expand the arsenal of health surveillance tools used to forecast seasonal flu and other viral respiratory outbreaks, such as the COVID-19 pandemic or SARS.

Models like these can be lifesavers by directly informing the public health response during a flu epidemic. These data sources can help determine the timing for flu vaccine campaigns, potential travel restrictions, the allocation of medical supplies and more.

Here’s how it works: a thermal camera detects the heat emitted from our bodies, and a microphone records sounds. That data is then fed into a Raspberry Pi and sent to a connected Intel Neural Compute USB Stick that runs a convolutional neural network model trained to recognize the sound of coughs. Thus the number of coughs can be counted, and plugged into forecast models along with the detected crowd size.

The model was trained on audio snippets of people coughing, sneezing, and clearing their throat from four different datasets. When the researchers deployed their system in four different waiting rooms at UMass's University Health Services clinic over seven months, FluSense collected 21 million audio samples and was accurate at detecting coughs 81 per cent of the time.


White House turns to Big Tech to fix coronavirus blunders while classifying previous conversations


"This may allow us to predict flu trends in a much more accurate manner," said Tauhidur Rahman, co-author of the paper and an assistant professor of computer and information sciences at UMass, on Thursday.

Andrew Lover, co-author of the paper and an assistant professor at the UMass’ School of Public Health & Health Sciences, wants to implement FluSense across more waiting rooms in clinics across the US to automate the measurement of people's coughing – one of the symptoms of the COVID-19 coronavirus among other afflictions. "We have the initial validation that the coughing indeed has a correlation with influenza-related illness," said Lover. "Now we want to validate it beyond this specific hospital setting and show that we can generalize across locations."

Sure, it's possible this kit can detect croaks to measure the COVID-19 spread, however, coughs from other stuff – like a lifetime of smoking, colds, and whatnot – are going to muddle the numbers, we suspect, given the accuracy rate. If was sophisticated enough to discern the type of cough, it may be more accurate in counting flu or COVID-19 splutters from common cold wheezes, if such a thing is possible.

The Register has asked the academics for extra explanation and comment. ®

More about

More about

More about


Send us news

Other stories you might like