Cardiac arrest often takes place at home, and people do not always have somebody around to ask for help. Your smart speaker may one day be helpful, although. Researchers at University of Washington have made a proof-of-concept AI that can identify the audio hints of cardiac arrest and answer for that reason. The system is educated on 911 call examples to listen for the telltale audios of agonal breathing that follow with cardiac arrest. It originally asks for assistance from individuals close by to offer CPR, but will dial 911 if there is no answer.
The training was decisively varied (such as by capturing recordings on various speakers and handsets) to stop as many fake positives as achievable, and the original accuracy seems to be very high. The AI only mis-detected breathing 0.22% of the time or less when it identified a single event, but it had perfect detection when it heard events minimum 10 Seconds apart.
This is not just a theoretical workout. The researchers have made a spinout firm (Sound Life Sciences) that aims to make this tech public. You may not see it emerge as a Google Assistant or Alexa feature considering the privacy implications, but it is not out of the world of possibility. It might make logic for dedicated devices and other health sensors, on the other hand. And no matter how it comes out, it can offer independence to at-risk people and seniors who need to know they will get assistance in a crisis.
On a related note, scientists at UC Berkeley and MIT have designed and now experiment with a device that can take out water out in the driest of climates from the air even. The team showed the device last year in an article and now they have enhanced the design and tested it out in Tempe, Arizona.