The overall goal of this proposal is to introduce a new paradigm for occupancy assessment, using ubiquitous wireless signals. Wireless signals are ubiquitous these days, which opens up the possibility of using them for sensing and learning about the environment. This work then develops the theoretical foundation and design tools for occupancy assessment with ubiquitous RF signals, with an emphasis on understanding the fundamental capabilities and limitations. This website will be updated throughout the course of this project.
Back to top
C. R. Karanam, B. Korany, and Y. Mostofi, "Tracking from One Side - Multi-Person Passive Tracking with WiFi Magnitude Measurements," in proceedings of the 18th ACM/IEEE International Conference on Information Processing in Sensor Networks (IPSN), April 2019 (acceptance rate: 27%).[pdf][bibtex]
B. Korany*, C. R. Karanam*, H. Cai*, and Y. Mostofi, "XModal-ID: Using WiFi for Through-Wall Person Identification from Candidate Video Footage," 25th Annual International Conference on Mobile Computing and Networking (MobiCom), Oct. 2019, accepted to appear (acceptance rate: 19%).[pdf][video] (*equal contribution)
S. Depatla and Y. Mostofi, "Occupancy Analytics in Retail Stores Using Wireless Signals," in proceedings of the IEEE International Conference on Sensing, Communication and Networking (Secon), June 2019 (acceptance rate: 28.6%).[pdf]
S. Depatla and Y. Mostofi, "Passive Crowd Speed Estimation in Adjacent Regions with Minimal WiFi Sensing," accepted to appear, IEEE Transactions on Mobile Computing, 2020.
Back to top
So far, we have proposed a new framework to estimate the occupancy dynamics over a large area, based on the received power measurements of wireless links that are sparsely installed throughout the space, and without relying on people to carry any device. We first prove how the cross-correlation and the probability of crossing the two links implicitly carry key information about the pedestrian speeds and develop a mathematical model to relate them to pedestrian speeds. We then exploit the sparsity in the spatial and temporal gradient of the occupancy dynamics and pose an optimization problem to estimate the arrival rates over the whole store, based only on a very small number of wireless measurements. We have thoroughly validated our framework with several experiments in three different retail stores – Kmart and two anonymous retail stores, using the RSSI measurements of Bluetooth Low Energy (BLE) Chips. Our results confirm that our framework can accurately estimate the rate of arrival of people into different aisles of a retail store with minimal wireless sensing.
Moreover, Mostofi’s lab has developed a new framework that has enabled, for the first time, identifying a person through wall from a candidate video footage, using only WiFi signals. Consider the case that a video footage of a person is available. A pair of WiFi transceivers are inserted outside of a building and are tasked with figuring out if the person in the video is behind these walls. We have shown, for the first time, that this is indeed possible. More specifically, we have proposed a framework that can translate the video content to the RF domain by using 3D mesh recovery algorithms and an efficient electromagnetic wave approximation on the extracted mesh. Then, our proposed signal processing pipeline extracts several key features from both the real RF signal, measured on the WiFI cards, and the video-to-RF one, and established if they belong to the same person. A video of this work went viral online (video link) and several reputable news agencies such as BBC covered the new invention.
Back to top