Description
Gait recognition identifies or verifies a walking subject from the periodic Doppler signature their stride imparts on a CSI stream. Unlike camera-based gait, the WiFi version sees through clothing and lighting, capturing torso-Doppler fundamental plus limb-Doppler harmonics. WifiU (Wang et al.) defined the canonical pipeline: bandpass filter → spectrogram → cycle extraction → classifier.
When it's used
- Identity-aware occupancy
- Authentication via behavioural biometrics
- Health monitoring (gait deterioration, Parkinsonian signatures)
- As a sub-task of
human-activity-recognition
Limitations
- Susceptible to walking-speed variation and footwear changes
- Limited robustness to direction of walk relative to TX/RX baseline
- Carry-state (bag, child) shifts the Doppler signature