Description
Estimating breathing rate (and increasingly tidal volume) of a stationary person from sub-millimeter chest-wall motion observable in WiFi CSI phase or mmWave radar range bins. Respiration is the canonical "fine-grained CSI sensing" problem because chest motion is at the very edge of what commodity hardware can resolve, and demonstrations of clean breathing-rate extraction validate the underlying signal model (Fresnel zones, CSI ratio).
Why it's hard
- Chest-wall motion is on the order of 1–5 mm; carrier wavelength is 5–6 cm at 5 GHz WiFi.
- SNR depends sensitively on subject location relative to Fresnel zone boundaries.
- Other periodic motions (typing, page-turning, fan rotation) contaminate the breathing band.
- Multi-person breathing is hard to separate without spatial diversity (multiple links / antennas).
- Posture and chest direction relative to link strongly modulate signal.
Common approaches
- Fresnel-zone modeling of CSI amplitude periodicity.
- CSI-ratio model to cancel time-varying carrier-frequency offset.
- Spectral-band filtering (0.1–0.5 Hz) plus peak detection.
- Deep models for multi-person separation and motion-artifact rejection.
Source Papers
- wu2022_75d3 ↗ — WiFi CSI device-free sensing: Fresnel zone to CSI-ratio model.
- wang2019_d6f9 ↗ — survey on human-behavior recognition using CSI (respiration covered).
- han2026_39eb ↗ — NearSense exploring NearLink for new-generation sensing.
- fallani2026_04be ↗ — IoT solutions for e-Health continuity at home.