Research On Wi-Fi-based Human Status Passive Sensing Algorithm For Indoor Health Monitoring

Under the background of the massive impact of global coronavirus epidemic,people are more concentrating on their physical and mental health.It not only makes people pursue a healthier lifestyle,but also promotes the research on health monitoring technology.With the development of wireless communication and sensing technology,the research of non-contact Wi-Fi sensing has attracted much attention over the past few years.Many researches have emerged in the application of indoor health monitoring.In this paper,by leveraging Wi-Fi physical layer parameter Channel State Information(CSI),we propose innovative passive Wi-Fi sensing algorithms to detect the fine-grained status(human respiration status)and the coarse-grained status(human walking trajectory)in indoor environment.Aiming at the fine-grained status of human respiration,we propose a human respiration status monitoring algorithm based on the mathematical model and machine learning algorithm of Wi-Fi sensing.The monitoring algorithm consists of human presence detection module and respiration status monitoring module.The former extracts the Doppler spectrum from CSI to classify human presence based on Naive Bayes classifier.To obtain detailed respiration status in the latter,we define the Respiration-to-Noise Ratio to select the most sensitive data streams.For detecting and distinguishing abnormal respiratory patterns,we extend the peak detection method and leverage machine learning based classifier in respiration apnea periods.We collect experimental data to conduct extensive experiments.The results demonstrate the classifier accuracy in the human presence detection module is above 93%and the correlation coefficient between the respiration status and the ground truth data is 80%.For the coarse-grained status of human walking trajectory,we propose a passive human tracking algorithm based on particle filter.Firstly,the CSI measurements are obtained,and the principal component is extracted by leveraging the correlation property of human walking motion on different Wi-Fi subcarriers.The Angle of Arrival(AoA)of human reflection signal is calculated based on the MUSIC algorithm after the spatial-temporal smoothing.Meanwhile,the complex phase measurements of the conjugate multiplication CSI on different receiver antennas are used to obtain the Doppler frequency shift.With the above two signal parameters,the nonlinear human walking state system model is designed,and the walking trajectory is estimated by particle filter.We build the prototyped system and experiments show that the accuracy of our approach on Wi-Fi parameter extraction is better than the comparing algorithm.The median tracking error of the proposed algorithm is 41.8cm,which is comparable to the state-of-the-art to achieve a decimeter-level tracking accuracy.