rTalk: Multi-Person Motion Tracking via RF Body Reflections

Meta note (or is it meta-note):

I have considered how best to provide short posts to get people interested in academic research that I find interesting. My past approach was to blast out variable length summaries that I titled reviews but did not really meet the qualification of a review. I wasn't assigning a score. Instead, I want to title these posts: "rTalk: <Title of Paper>." The idea of these posts is to simply talk about the paper. A summary with some intuition if you will.

I thank you, the reader, for your persistence with respect to this blog's growing pains. It's a dynamic system of maturity and growth, for both the blog and myself.

---

Multi-Person Motion Tracking via RF Body Reflections
Fadel Adib, Zachary Kabelac, and Dina Katabi
download paper

The authors present WiZ, a multi-person, centimeter scale motion tracking system. This system can spatially localize humans, two at most with the current implementation, and track hand gestures using radio reflections in the presence of obstructions such as walls (i.e., static objects). The authors use the time-of-flight (TOF), or time of a directed RF signal to leave the transceiver and return to a receiver via reflection, to track motion. Collected TOFs are composed into a heatmap visualization that presents reflections color-coded from red to blue. Red colors, or hot, indicate areas where reflections occurred --- we found a person! Blue colors, or not-so-hot, indicates areas of little reflection --- its just a wall. Specifically, we would imagine a human body on a two dimensional plane as being red and the emptiness around her as blue. This technique, however, is non-optimal, or simply really difficult, when tracking more than one person. The intuition as to why is because received signal reflections may overpower others (e.g., a person standing further away) and there is no way to distinguish individual reflected signal points (e.g., which signal was reflected from person A).

 The authors introduce two new algorithms to overcome this limitation. They are successive silhouette cancelation (SSC) and multi-shift frequency modulation carrier (FMCW) waves. Multi-shift FMCW delays transmit signals for some time equal to the TOF limit, or maximum time a TOF could take. This allows us to distinguish received signal reflections from one another and thus use multiple transmitters and receivers. Successive silhouette cancelation addresses what is informally referenced as the near/far problem. This problem is described in the example above as two people which we would like to track, however one person is closer to the receiver and thus overwhelms the received signal reflection from the person that is further away. The authors propose SSC as a series of algorithmic steps to first detect the strongest reflector (i.e., the person that is closer), remap the reflectors locations to the set of TOFs, cancel out the ramped TOFs, and finally re-compose the heatmap visualization. This process can be iterative, thus tracking motion of individuals exceedingly further away.

If you're interested in seeing the experimental setup and a different summarization, please check out the following references: http://www.extremetech.com/extreme/184347-mit-perfects-cheap-accurate-through-wall-movement-and-heartbeat-detection-with-wifi and http://newsoffice.mit.edu/2014/could-wireless-replace-wearables. Before closing, I want you to think about how new technologies, such as motion tracking here, might impact privacy. Are the implications similar to Kyllo v. United States? How could privacy advocates protect themselves? If deployed, what sort of companies might use this technology (e.g., wifi tracking via smartphone, or iBeacon). What positives results are there for motion tracking. Fantastic work overall. I, personally, cannot wait to see what this group does next!