Radio technology has been widely used for high-speed wireless communications. In the near future, radio technology would provide sensing capabilities to enable a diversified indoor applications in the era of Internet of Things (IoT). This is because that the electromagnetic (EM) wave, emitted from the transmitter propagates through multipath before arriving at the receiver, is varied by the environmental perturbations. Such variations in EM waves reveal important environmental changes useful for IoT applications. Thus, in IoT networks, radios are not only the ubiquitous communication interfaces but also exhibit augmented sensing potential.

Despite the wide variety of IoT devices, most of them are equipped with WiFi which is a very mature and cost-effective connectivity solution and has evolved significantly ever since its standardization. Meanwhile, as people are spending more and more time indoors, most indoor spaces have been already equipped with WiFi infrastructures, which makes the IoT devices empowered by WiFi to blend into the existing WiFi infrastructures without efforts. Therefore, it is highly valuable to adopt radio analytics to analyze the WiFi radio signals to facilitate key IoT applications.

In this dissertation, we explore the viability of using WiFi for two important IoT applications: indoor localization and vital sign monitoring. In the first part, we propose two indoor localization systems (IPSs) leveraging the time-reversal (TR) technique on off-the-shelf WiFi devices. The proposed IPSs utilize the location-specific features, i.e., the channel frequency response (CFR), which is a fine-grained information readily available on off-the-shelf devices that depicts the propagation of EM waves from the transmitter to different locations. The proposed IPSs consist of an offline phase which collects CFRs from locations-of-interest, and an online phase which compares the instantaneous CFRs with those captured in the offline phase. To calculate the similarities among locations, the TR focusing effect is evaluated quantitively between each pair of CFRs associated with these locations using the TR resonating strength (TRRS). Realizing that the bandwidth limit on mainstream WiFi devices could lead to location ambiguity, we exploit two diversities inherent in WiFi devices, i.e., frequency diversity and spatial diversity, to expand the effective bandwidth. Extensive experiments show a localization accuracy of 1 to 2 centimeters even under strong non-line-of-sight (NLOS) conditions as well as enhanced robustness against environmental dynamics.

In the second part, we investigate the feasibility of high accuracy vital sign monitoring using CFRs. First of all, we present a highly accurate breathing monitoring system. Realizing that breathing injects tiny but periodic signals into the WiFi signal, we project the CFR time series onto the TRRS feature space to amplify such CFR perturbations. Integrated with machine learning techniques, the proposed scheme could distinguish breathing rates associated with different people. In addition, it could detect the presence of breathing and count the number of people. The performance is demonstrated by extensive experiments in multiple environments. Secondly, we present a lightweight vital sign monitoring solution with a much reduced computational complexity. Moreover, we supplement the proposed vital sign monitoring system with a finite state machine (FSM) to remedy the impact of motions on the monitoring performance. Extensive experimental results demonstrate the excellent performance of both breathing monitoring schemes.