Computational Wireless Sensing at Scale

Abstract

Computational wireless sensing is an exciting field of research where wireless signals from every-day computing devices are used to enable sensing. The key challenge is to enable new sensing capabilities that can be deployed at scale and have an impact in the real world. In this dissertation, we show, for the first time, how to enable computational wireless sensing at scale by leveraging ubiquitous devices like smartphones. We present algorithms that can wirelessly sense motion and physiological signals such as breathing using just a smartphone, in a contactless manner. The key idea in this dissertation is to transform the smart devices into active sonar systems. We emit an in-audible sound signal from the device’s speaker and these signals are reflected back by the different objects and people in the environment which we then capture using the device’s microphones. By isolating the various reflections at the receiver, we can enable the sensing and tracking of different objects in the environment. Building on this approach, we design, build and evaluate four key systems. We present the first contactless system that can diagnose sleep apnea using a smartphone by monitoring the minute millimeter level breathing motion. Next, we built a smartphone based system that can detect opioid overdoses by looking at respiratory depression and body movements. We then show that, beyond respiratory tracking, we can also track the 2D location of a finger around a smartphone or watch to enable new interaction modalities. Finally, we explore the privacy implications of enabling wireless sensing on commodity devices and propose defenses to prevent information leakage