A research paper presents Wi‑Mesh, a system that uses Wi‑Fi signal reflections to construct detailed 3D human body meshes without cameras or wearable sensors. By leveraging multiple antennas on standard Wi‑Fi devices, the approach estimates the directional properties of reflected signals and converts them into visual representations of the human body shape and motion.
The system works by estimating two‑dimensional angles of arrival from Wi‑Fi reflections to generate image‑like representations of the body, then applies deep learning models to transform these into full 3D mesh reconstructions. In tests, the method achieved positional accuracy for body joints and surface vertices that is comparable to systems requiring specialized hardware.
Top comments (0)