Duet: Estimating User Position and Identity in Smart Homes Using Intermittent and Incomplete RF-Data

AbstractAlthough past work on RF-based indoor localization has delivered important advances, it typically makes assumptions that hinder its adoption in smart home applications. Most localization systems assume that users carry their phones on them at home, an assumption that has been proven highly inaccurate in past measurements. The few localization systems that do not require the user to carry a device on her, cannot tell the identity of the person; yet identification is essential to most smart home applications. This paper focuses on addressing these issues so that smart homes can benefit from recent advances in indoor localization.We introduce Duet, a multi-modal system that takes as input measurements from both device-based and device-free localization. Duet introduces a new framework that combines probabilistic inference with first order logic to reason about the users' most likely locations and identities in light of the measurements. We implement Duet and compare it with a baseline that uses state-of-art WiFi-based localization. The results of two weeks of monitoring in two smart environments show that Duet accurately localizes and identifies the users for 94% and 96% of the time in the two places. In contrast, the baseline is accurate 17% and 42% respectively.