Together or Alone: Detecting Group Mobility with Wireless Fingerprints Gürkan SOLMAZ and Fang-Jing WU NEC Laboratories Europe, CSST group, Heidelberg, Germany 24 May 2017 This work has received funding from the European Union s Horizon 2020 research and innovation programme within the project Worldwide Interoperability for Semantics IoT under the grant agreement No 723156. Responsibility for the information and views set out in this document lies entirely with the authors.
Motivation Idea: Detecting people that walk together OR alone (group vs. individual mobility) using wireless signals captured from mobile devices A flexible approach, based on a novel perspective Different than previous camera-based or smartphone app-based approaches Does not require priori knowledge (real-time detection) Does not rely on localization, applicable to indoor and outdoor environments Applicable for various conditions Easy to setup system (Multiple Raspberry PIs deployed in the targeted areas) Darkness, blind spots, behind the walls Low computation & communication overhead Possible applications: Short term: Profiling and surveillance Long term: Identify social interactions in the crowds (e.g., university campus) 2 NEC Europe Ltd. 2016
Targeted environments The idea is applicable to both indoor and outdoor environments Initially we targeted office environment for easy & longer term experimenting Possible outdoor environments: Streets, city squares Leisure areas (parks in a city) Ski resort Theme parks Indoor environments: Airports Shopping malls Train stations Stadiums 3 NEC Europe Ltd. 2016
System design (1/2) Easy to setup system Can be fed with extra learning (e.g., by camera) for calibration in specific environments 4 NEC Europe Ltd. 2016
System design (2/2) Back-end server Network gateway Wireless sniffers Beacons carried by humans We implemented a beacon based sniffing system as a prototype for opt-in data collection Design is applicable to the Wi-Fi-based solutions 5 NEC Europe Ltd. 2016
Method (1/2) Raw data Aggregated (5s) Extracted signals 6 NEC Europe Ltd. 2016
Method (2/2) Movement status (Dynamic vs. Static) based on sniffer fingerprints S ordered list of sniffers for person P at time interval T f outputs the first k element of the list (1 k n, n number of sniffers) Space correlation between two people P i, P j Dynamic status & correlation group mobility! 7 NEC Europe Ltd. 2016
Experimental setup Controlled experiment for collecting ground truth 10 beacons carried together Real-world experiment in the office environment Bluetooth data collected from 10 participants for 2 weeks 2 participants carried 2 beacons at the same time for reliability & consistency 8 NEC Europe Ltd. 2016
Experimental study controlled experiment Beacon 1 Beacon 4 Beacon 9 Aggregated results from 3 randomly selected beacons All pairs have on average (99.4%) similarity to each other Wireless fingerprints similarity 9 NEC Europe Ltd. 2016
Experimental study group mobility detection 10 beacons 1 day 2 weeks 4 group movements detected for all 10 beacons 300 alone walks, 20 walks by groups of size 2 in one day Group sizes of 2, 3, 4, 5, and 6 About 2 minutes needed to compute one day 10 NEC Europe Ltd. 2016
Experimental study reflections Movement intersections (incl. alone) Movement intersections (excl. alone) People working in the same office room tend to walk together Only exception: External member P 6 tends to walk with P 1, P 2, P 3 (e.g., going to lunch together) Results of the 2 beacons On average 95.39% similarity score, %79 movement intersection 11 NEC Europe Ltd. 2016
Questions? gurkan.solmaz@neclab.eu This work has received funding from the European Union s Horizon 2020 research and innovation programme within the project Worldwide Interoperability for Semantics IoT under the grant agreement No 723156. Responsibility for the information and views set out in this document lies entirely with the authors. 12 NEC Europe Ltd. 2016