Exploiting Dynamic Phase Information for Respiration Monitoring During Sleep via WiFi

Respiration monitoring especially during sleep is essential for diagnosing the symptoms of sleep-related diseases. Recently, compared to the conventional wearable devices, WiFi signals can provide a better sensing way with comfortable and convenient benefits. However, the static components caused by static objects closed to WiFi devices significantly limit the detection performance of the respiration sensing. To address this issue, by exploiting the dynamic phase information of WiFi channel state information, we propose a method for extracting the respiration signal. Specifically, we first use the nonlinear least squares fitting to eliminate the static component in WiFi signals. Then, we use principal component analysis to extract the respiration-related dominant component to against the relatively large noise, i.e., the low sensing-signal-to-noise ratio. Our approach has been evaluated on a dataset of 11 participants. Results indicate that it outperforms existing state-of-the-arts in estimating respiration rate, achieving a mean absolute error of less than 0.3 bpm (breaths per minute) across sleep postures.