A study demonstrated this using an AI and two commercial WiFi routers.
There are different ways to situate someone in an environment. Cameras, microphones and other miscellaneous sensors are widespread and their concepts are often quite simple to understand. However, human ingenuity seems like a bottomless pit, and researchers from Carnegie-Mellon University in Pennsylvania are proving it again.
Highly accurate wireframe image
By placing two routers at each end of a room, TP-Link Archer A7 AC1750 to be precise, it was possible to determine, and even visualize, the presence, or not, of a variable number of people in the latter. . By comparing the transmitted signal with the received signal, a neural network manages to make an estimate of the observed environment. Then, by breaking down the human forms into 24 segments, the researchers were able to get a fairly clear picture of the position, or even the pose, of the individuals present.
However, the system is not infallible. Since the data must be processed by an artificial intelligence, the results can be affected by a certain “subjectivity” of the latter, which is not yet able to recognize a person with an unusual pose. Additionally, it has been found that too many individuals in a room can affect the results. But, unlike technologies capable of obtaining results of the same type, such as radar or LiDAR, this technique is less expensive in terms of energy and equipment, since the routers used only cost a few tens of dollars.
An inexpensive technique that would have its advantages… and its disadvantages.
For the authors of the study, there is still a lot of work to be done to make this technique viable. They believe that it will be necessary to train a dedicated AI, the one they used being rather trained to recognize human subjects on digital images. This should be able to work in different room configurations, recognize smaller groups of people, and more easily visualize unusual poses and situations that are currently lacking.
If it is possible to imagine many applications for such an accessible tool, it is for the moment mainly presented as a means of controlling the safety of the elderly and living alone. Indeed, the cameras can be very visible and can quickly become annoying when other people have access to them. In addition, they are generally not very effective in the dark and are very sensitive to obstructions, which is not the case with wifi devices as long as one can control the various interferences. However, if respect for privacy is put forward by researchers, one could also conceive of the opposite.
Source : Tom’s Hardware, Carnegie Mellon University
6