ECE Seminar: Human Sensing Using Light
Monday, March 4, 2019 - 12:00pm to 1:00pm
Tianxing Li, PhD candidate in Computer Science, Dartmouth College
Long-term, continuous monitoring of human behaviors and biological markers provides essential input for data-driven health technologies. Existing sensing systems, however, still have significant drawbacks. Wearable and wellness sensors (e.g., Apple Watch, Fitbit) require frequent charging and impose extra burdens on users, while camera-equipped systems present serious privacy risks by capturing raw images. Tackling these challenges demands novel system designs that can strike a better balance between sensing granularity, power consumption, and privacy. In this talk, I will describe our recent efforts to enable a variety of sensing capabilities with minimal, low-level sensing data. Specifically, we have explored the use of light (visible or near infrared light) as the sensing medium and studied how the human body interacts with light to infer fine-grained behaviors and biomarkers. I will first present my work on reconstructing skeleton poses. Existing methods all require invasive cameras with a limited field of views. My work replaces cameras with photodiodes to capture how the user body blocks light and aggregates such binary blockage information observed in different directions to reconstruct skeleton poses. I will then describe our development of the first battery-free eye tracker without the need of cameras.