Like all radio waves, Wi-Fi signals undergo subtle shifts when they encounter objects—human beings included. These can reveal information about the shape and motion of what has been encountered, in a manner akin to the way a bat’s chirps reveal obstacles and prey.
Starting from this premise Jiaqi Geng, Dong Huang and Fernando De la Torre, of Carnegie Mellon University, in Pittsburgh, wondered if they could use Wi-Fi to record the behaviour of people inside otherwise unobservable rooms. As they describe in a posting on arXiv, they have found that they can. “DensePose from Wi-Fi”, the paper in question, describes how they ran Wi-Fi signals from a room with appropriate routers in it through an artificial-intelligence algorithm trained on signals from people engaging in various, known activities. This algorithm was able to reconstruct moving digital portraits, called pose estimations, of the individuals in the room.
No comments:
Post a Comment