Researchers create AI microphone that listens to coughing and sneezing in public spaces to predict how many people have respiratory illnesses at one time, even if they haven’t seen a doctor for treatment
- Researchers have created a device that analyzes audio samples to detect cough
- The device, called FluSense, has AI that can estimate the percentage of people in public space who suffer from a sound-based respiratory disease
- After initial tests in four clinic waiting rooms, they discovered that his predictions were “strongly correlated” with the actual diagnoses of the doctors in those clinics.
Researchers at the University of Massachusetts at Amherst have created an AI that listens to coughing and sneezing sounds to estimate the percentage of people in public space who have respiratory illness.
The device, called FluSense, was initially tested over an eight-month period in four clinic waiting rooms on the university campus.
In addition to recording “non-vocal” audio samples, FluSense is also equipped with a thermal camera to search for people at high temperatures.
Scientists at the University of Massachusetts at Amherst have created an AI powered device called FluSense (pictured above) to analyze audio samples from public spaces to estimate the percentage of the population suffering from respiratory disease based on coughing, sneezing and body temperature readings
According to its co-creator, Tauhidur Rahman, the device is not intended to distinguish individual cases of disease but to capture trends in the population to see if something is developing that may not yet have been detected in medical tests.
“ I thought if we could capture coughing or sneezing sounds in public spaces where a lot of people gather naturally, we could use this information as a new data source to predict epidemiological trends, ” a- he declared. Told UMass Amherst’s news blog.
The first tests of the device took place between December 2018 and July 2019, and the team was initially interested in predicting the potential spread of seasonal flu and other respiratory illnesses.
During the test window, FluSense devices analyzed more than 21 million non-voice audio samples and 350,000 thermal images.
The AI used the audio samples to estimate the size of the populations in the different waiting rooms, then calculated the percentage of people likely to have had a respiratory illness according to the frequency of coughing, sneezing and signatures high temperature.
The FlueSense prototype is manufactured using a Raspberry Pi and fits into a thick casing of a large dictionary, and was initially tested in four clinic waiting rooms on the UMass Amherst campus, where his estimates were strongly correlated with actual clinical diagnoses
The team is preparing for the next phase of testing, moving the device to other larger public spaces to see how useful they can be in modeling the percentage of the population ill at any given time, regardless of whether whether they were the doctor or not
The team compared FluSense predictions with laboratory results from clinics and found that they were “highly correlated” with actual disease levels.
Building on its initial success, the team plans to extend the tests to other public places outside the clinics.
The devices are made with a Raspberry Pi, a network of microphones and a thermal imaging camera, all assembled in a box the size of a small dictionary, which makes them relatively inexpensive to assemble and distribute.
“We have initial validation that coughing does correlate with flu-related illness,” said Andrew Lover of UMass Amherst.
“Now we want to validate it beyond this specific hospital framework and show that we can generalize across locations.”