Continuous monitoring of vital signs (e.g., respiration and heart rate, heart rate variability, etc.) is critical for early detection and prevention of potentially fatal diseases. Existing solutions usually require users to wear dedicated devices such as wrist-worn sensors or chest straps to physically contact with human body, which is uncomfortable for users and sometimes may cause skin allergies. With the rapid development of the Internet of Things (IoT), wireless sensing has gained increasing attention in recent years because of the ubiquitous deployment of wireless devices. It has been proved that the presence of human will affect wireless signal propagation, enabling the functionality of wirelessly monitoring human subjects by analyzing the electromagnetic (EM) wave.

Despite of the wide variety of IoT devices, most of them are equipped with WiFi, which is a very mature and cost-effective connectivity solution. Moreover, as the next-generation wireless communication technique, millimeter wave (mmWave) radio has become available on home routers, vehicles, etc., to achieve higher performance (e.g., larger bandwidth, higher directionality). Motivated by the increasing demand of monitoring vital signs as well as the development of IoT, in this dissertation, we propose four wireless sensing systems to monitorvital signs leveraging the channel information of commercial devices.

In the first part, using the Channel State Information (CSI) of a single pair of commercial WiFi devices, a novel system is proposed to continuously track the breathing rates of multiple persons without requiring prior knowledge of crowd number. By leveraging both the spectral and temporal diversity of the CSI, the proposed system can correctly extract the breathing rate traces of multiple users even if some of them merge together for a short time period. Furthermore, byutilizing the breathing traces obtained, the crowd number can be estimated for the occupancy level estimation in the smart home or smart office scenario.

In the second part, we propose a multi-person Respiration Rate (RR) as well as Heart Rate (HR) monitoring system leveraging the Channel Impulse Response (CIR) of a 60GHz WiFi. A calibration-free object detector is first designed to identify static objects, stationary human subjects and human in motion using both the amplitude and phase of the CIR measurement. To get robust HR estimations corresponding to stationary human subjects, the respiration signal is first eliminated from the phase of the CIR measurement before obtaining the spectrogram of heartbeat signal. Dynamic programming is further adopted to get the final estimation of HR by exploiting both the temporal and spectral information. Experimental results demonstrate promising performance of the proposed system, including the Non-Line-of-Sight (NLOS) scenario.

To further get finer information of heartbeat signal, in the third part, we propose mmHRV, the first multi-user Heart Rate Variability (HRV) monitoring system, using a commercial mmWave Frequency-Modulated-Continuous-Wave (FMCW) radar. We first develop a calibration-free target detector to identify the number of users and their locations. Then the heartbeat signal of each user is obtained by optimizing the decomposition of the composite phase measurement modulated by the chest movement. The exact time of heartbeats are estimated by identifying the peak location of the estimated heartbeat signal, and Inter-Beat Intervals (IBI) can be further derived to evaluate HRV. Extensive experiments have been conducted to explore the influence of different settings, including the distance between human and device, user orientation, incidental angle and NLOS setting, etc..

In the final part of this dissertation, we propose a driver vital sign monitoring system built upon a commercial FMCW radar. The system first eliminates driver’s motion artifacts by a two-step motion compensation module. Then the respiration and heartbeat signals are estimated simultaneously by jointly decomposing the phase measurement over all range-azimuth bins containing vital signals. The RR, HR and IBI are further derived using the estimated respiration and heartbeat signals. We evaluate the system performance in real driving environment, where the impact of pavement condition, device location as well as motion type are explored in the experiment.