Smart Speakers May Soon be Able to Tell if You’re Having a Heart Attack
Cardiac arrest caused the deaths of more than 350,000 Americans in 2015 alone. Often time, those who experience cardiac arrest do not have anyone near the vicinity to call for help, making it difficult for emergency services to arrive on time to provide aid. However, your smart speaker may one day be able to help in these situations.
University of Washington researchers created a proof-of-concept AI that is fined tuned to detect the specific audio cues of someone who is going into cardiac arrest. The AI is trained using machine learning by being given 911 call samples of actual victims experiencing cardiac arrest and listening to the telltale sounds of gasps of air, also known as agonal breathing. If it detects these signs, it will ask the victim if there is anyone in the area to provide CPR. If no response is given within a short time frame, then it will call 911.
According to the research, the AI was exposed to a variety of different recordings from different phones and speakers of varying sound quality. This was to ensure that the AI would be trained well enough to avoid false positives of a person going into cardiac arrest. Initial accuracy of the testing revealed that it was fairly accurate, only misidentifying agonal breathing 0.22 % of the time. This accuracy was flawless when it listened to a recording for more than 10 seconds.
This technology is looking to be commercialized by these scientists, who have already created a company called Sound Life Sciences. Whether it will show up on your Google Home or Amazon Echo remains in question due to privacy concerns, there is potential for a new dedicated device to hit the market in the near future that will provide these protective services to customers. It could have major implications for seniors or at-risk people who do not have help at home.