Demo: Fusing WiFi and Video Sensing for Accurate Group Detection in Indoor Spaces

Abstract

Understanding one's group context in indoor spaces is useful for many reasons -- e.g., at a shopping mall, knowing a customer's group context can help in offering context-specific incentives, or estimating taxi demand for customers exiting the mall. We presented GruMon in Sen et. al, which detects groups accurately under accurate localization assumptions or with the availability of inertial sensors from smartphones carried by the users. However, in most real situations, (1) client-side sensory information is not available, and (2) server-side localization is erroneous. Further, detecting people who not carry mobile phones with them (e.g., children/elders and phones with WiFi turned OFF, which we refer to as ``hidden nodes"), and separating out smaller sub-groups from a larger group that happens to share similar trajectories, remains a challenge. In this demo paper, we present our improved system for group detection in indoor environments. We overcome the key challenges of localization errors, presence of hidden nodes and sub-groups by fusing the WiFi and video sensing modalities.