Low-Velocity CSI-Based Wi-Fi Sensing for Device-Free Respiratory Monitoring

Abstract

Wi-Fi sensing enables low-cost and device-free monitoring of subtle human motion. However, respiration-induced chest movement is weak and slow, making Doppler extraction challenging under low-SNR and non-line-of-sight (NLOS) conditions. This paper presents a Wi-Fi CSI-based framework for reconstructing low-velocity respiratory dynamics. The proposed method uses cross-antenna CSI ratios to suppress hardware-induced phase distortion, followed by eigen-beamforming and reassigned time–frequency analysis to enhance weak Doppler signatures and estimate motion velocity. Validation using a programmable linear guide with laser ToF reference showed 0.02 cm/s velocity error and 1 mm/s motion sensitivity. Human experiments further showed that the Wi-Fi-derived trajectory was consistent with ECG-derived respiration and RR-interval references, demonstrating its feasibility for cycle-level respiratory monitoring.