WiFi sensing method separates motion speed from location effects
Untangling the Geometry and Speed for RF Sensing Spectrograms
Machine Learning
Summary
WiFi signals can be used to sense movement, but the signals often mix up how fast something moves with where it is. The authors created a new way to separate these two effects so that the sensing is clearer and more accurate. They developed a model that looks at WiFi signal patterns and can figure out both the speed and position of a moving object. Their system was tested with many real and simulated examples and outperformed older methods.
What this means in practice
- •For smart home device makers: Improve in-home activity tracking by accurately distinguishing speed and position of moving people using WiFi signals without extra sensors.$Commercial implications: Enables better non-intrusive monitoring products that sell to consumers wanting smart home activity awareness.
- •For building security teams: Detect and track moving targets inside buildings more reliably by separating motion speed from location influences in WiFi signals.
Authors
Mert Torun, Darius Cuenca, Yasamin Mostofi
Abstract
A fundamental challenge in RF sensing is that Doppler signatures observed by a link entangle the target's motion with the sensing geometry, resulting in limited applicability to unconstrained real-world settings. In this paper, we establish a new foundation for physically interpretable RF sensing that disentangles reflector speed from geometry, jointly recovering the speed, geometry factor, relative amplitude, and width of each dominant Doppler ridge. More specifically, we first develop a compact parametric representation of WiFi spectrograms and establish its low-dimensional structure through a systematic computer-vision analysis of a large and diverse human-activity dataset, thereby providing a tractable foundation for learning. Building on this representation, we then design a physics-informed autoencoder whose structured bottleneck and differentiable RF forward model enforce physically meaningful estimates of reflector speed and geometry. We further introduce a synthetic-to-real training framework, eliminating the need for real WiFi training data. We extensively validate the proposed framework under both known and time-varying geometries, using both independently generated synthetic test sets and 31 real WiFi experiments. The results demonstrate the superior performance in speed and geometry extraction, robustly recovering the underlying geometry, speeds, Doppler-ridge amplitudes, and ridge widths across all settings, while substantially outperforming the strongest baselines.