With the rapid development of the Internet of Things (IoT), wireless sensing has found wide applications from wellbeing monitoring, activity recognition, to indoor tracking. In this dissertation, we investigate the problem of wireless sensing for IoT applications using only ambient radio signals, e.g., WiFi, LTE, and 5G. In particular, our work mainly focuses on passive speed estimation, motion detection, sleep monitoring, and indoor tracking for wireless sensing.

In this dissertation, we first study the problem of indoor speed estimation using WiFi channel state information (CSI). We develop the statistical electromagnetic (EM) wave theory for wireless sensing and establish a link between the autocorrelation function (ACF) of the physical layer CSI and the speed of a moving object. Based on the developed statistical EM wave theory for wireless sensing, we propose a universal low-complexity indoor speed estimation system leveraging CSI, which can work in both device-free and device-based situations. The proposed speed estimator differs from the other schemes requiring strong line-of-sight conditions between the source and observer in that it embraces the rich-scattering environment typical for indoors to facilitate highly accurate speed estimation. Moreover, as a calibration-free system, it saves the users' efforts from large-scale training and fine-tuning of system parameters. The proposed speed estimator can enable many IoT applications, e.g., gait monitoring, fall detection, and activity recognition.

Then, we also study the problem of indoor motion detection using CSI. The statistical behaviors of the CSI dynamics when motion presents can be characterized by the developed statistical EM theory for wireless sensing. We formulate the motion detection problem as a hypothesis testing problem and also derive the relationship between the detection rate and false alarm rate for motion detection, which is independent of locations, environments and motion types. Thus, the proposed motion detection system can work in most indoor environments, without any scenario-tailored training efforts. Extensive experiments conducted in several facilities show that the proposed system can achieve better detection performance compared to the existing CSI-based motion detection systems while maintaining a much larger coverage and a much lower false alarm rate.

This dissertation also focuses on sleep monitoring using CSI. First, we build a statistical model for maximizing the signal-to-noise (SNR) ratio of breathing signal, which accounts for all reflecting and scattering multipaths, allowing highly accurate and instantaneous breathing estimation with best-ever performance achieved on commodity devices. Our results demonstrate that the proposed breathing estimator yields a median absolute error of 0.47 bpm and a 95%-tile error of only 2.92 bpm for breathing estimation, and detects breathing robustly even when a person is 10m away from the WiFi link, or behind a wall. Then, we apply machine learning algorithms on the extracted features from the estimated breathing rates to classify different sleep stages, including wake, rapid eye movement (REM), and non-REM (NREM), which was previously only possible with dedicated hardware. Experimental results show that the proposed sleep monitoring system achieves sleep staging accuracy of 88%, outperforming advanced solutions using contact sensor or radar.

The last work of this dissertation considers the problem of indoor tracking using CSI. First, we leverage a stationary and location-independent property of the time-reversal (TR) focusing effect of radio signals for highly accurate moving distance estimation, which plays a key role in the proposed indoor tracking system. Together with the direction estimation based on inertial measurement unit and location correction using the constraints from the floorplan, the proposed indoor tracking system is shown to be able to track a moving object with decimeter-level accuracy in different environments.