Researchers at the U.S. National Institute of Standards and Technology (NIST) разработали AI algorithm BreatheSmart to detect breathing problems from subtle changes in Wi‑Fi signals.
According to the researchers, in laboratory conditions the deep learning model achieved an accuracy of 99.54%.
Wi‑Fi signals are almost everywhere. They bounce and pass through surfaces, attempting to link devices to routers. Any human actions, including the movements of their bodies during inhalation and exhalation, alter the propagation of the signal.
The human chest operates differently during quiet breathing and coughing, and differences in Wi‑Fi propagation can be traced.
Other researchers have studied using wireless signals to detect people and recognise their movements. However, these approaches required specialised sensing devices and large datasets.
In 2017 Origin Wireless представила an algorithm that uses a wireless mesh network to detect even minute movements. According to the company, the technology works with any compatible equipment.
NIST said that BreatheSmart can also be used with routers and devices already available on the market. For the system to work, only one router and a connected device are required.
The researchers modified the router’s firmware so that it checks the channel state information (CSI) ten times per second. These are signals sent by devices such as a smartphone or laptop to the router.
The router will detect changes in case of deviations in the surrounding environment.
The researchers simulated several variants of human breathing using a mannequin. They tracked changes in CSI signals through the router and a receiving device.
To make sense of the collected data, the scientists developed BreatheSmart.
According to the specialists, their solution will help third-party developers create software for remote monitoring of a person’s breathing using existing equipment.
“We collect data using software on the router. But this can be done via a mobile app,” said NIST scientist Jason Koder.
In September, American researchers announced the creation of an AI system for diagnostics and study of diseases by voice.
Subscribe to ForkLog news on Telegram: ForkLog AI — all the news from the AI world!