Scientists from the University of Carnegie Mellon have introduced an innovative method for determining the location and pose of people indoors using ordinary home Wi-Fi technology. This new approach offers advantages over traditional methods like video surveillance or lidars, as Wi-Fi is widely available and can collect data even in low-light conditions with obstacles present.
The principle of operation of the method
The basis for this method was the model Densepose, originally designed for analyzing human poses in photographs. Densepose identifies the human body in an image, divides it into zones corresponding to different body parts, and analyzes each zone separately. This enables accurate determination of people’s poses even when visual information is complex, such as overlapping figures or motion.
Researchers adapted this technology for Wi-Fi signals, converting analog signals from routers into data comparable to Densepose models. The concept is that changes in Wi-Fi signals due to human body movement and position can be interpreted similarly to changes in images.
To test their hypothesis, scientists set up a test scene with two Wi-Fi routers—one as a transmitter and the other as a receiver. A camera was also installed next to the receiver for comparison with visual data received through Wi-Fi.
General scheme of the test stand
The results of the experiments
Experiments demonstrated that the Wi-Fi-Densepose model can accurately recognize people’s positions and poses based on Wi-Fi signal analysis. However, the method’s accuracy is lower than traditional recognition methods, especially in complex poses or multi-person scenarios.
Restrictions and prospects
Despite its potential, the method has limitations such as signal interpretation challenges, low accuracy in crowded scenes, and possible interference, making real-world implementation difficult. The technology struggles with non-standard poses and multiple bodies in the same space.
While the University of Carnegie Mellon’s development presents new opportunities for monitoring and analyzing people’s movements without optical equipment, practical implementation requires addressing technical and methodological issues.