Smart speakers might one day be able to do more than read out the weather forecast and set cooking timers in the kitchen – they could also save your life.
Researchers from the University of Washington have developed a machine learning algorithm that detects the sounds of agonal breathing, the sharp gasping for air that accompanies around 50 percent of cardiac arrest events.
If the tool were built into a smart speaker such as an Amazon Echo or a Google Home, the researchers say, it could sense someone having a heart attack and raise the alarm, even if the person involved is incapacitated.
"We envision a contactless system that works by continuously and passively monitoring the bedroom for an agonal breathing event, and alerts anyone nearby to come provide CPR," says one of the team, Shyam Gollakota. "And then if there's no response, the device can automatically call 911."
More data needed
The system isn't ready for primetime yet though. The algorithm needs to be trained on a broader range of data to reduce false positives and to make sure agonal breathing can be detected across a wider distance and from different angles.
Audio from real 911 calls in Seattle was used to train the AI running the system, as well as audio sampled from sleep studies. The software ended up correctly detecting agonal breathing 97 percent of the time at a distance of up to 6 meters, which is a good start.
There were some false positives flagged up during testing, however, and the team wants to make sure the underlying algorithm is as accurate as it can be before it's considered for use in the real world.
We may well see more smart speaker applications like this in the future: Alexa can already listen out for the sound of glass breaking and alert you of a potential break-in.