Detecting and Understanding Human Movement With Mobile Devices Through WiFi-based Indoor Localization Technology

Mr Hong Hande
Dr Chan Mun Choon, Associate Professor, School of Computing

  19 Jun 2018 Tuesday, 10:30 AM to 12:00 PM

 Executive Classroom, COM2-04-02


Human movement and location detection have long been important and active research areas that have attracted extensive efforts from the research community. Human activity detection generally encompasses a range of topics including detecting the presence of human bodies, crowd moving pattern and social relationship between individuals. The proliferation of WiFi-enabled devices that are constantly tethered to users provide good proxies to reason about a user's location, behavior, and social interaction pattern. Thus, the use of WiFi related information to extract context information become an effective tool to understand human movement and behavior. In this thesis, we present three closely related works making use of this information.

We first present SocialProbe, a system to extract social behavior and interaction patterns of mobile users by passively monitoring WiFi probe requests and null data frames that are sent by smartphones for network control/management purposes. By analyzing the temporal and spatial correlations of the RSS of packets from these low rate transmissions, we are able to discover proximity relationships, occupancy patterns, and social interactions among users.

SocialProbe focuses more on the long-term behavior of people in a specific location. The second piece of work, CrowdProbe applies the passive monitoring technique to track crowd movement between different locations. We propose a Hidden Markov Models (HMM) based visitor trajectory inferring method based on passive WiFi monitoring. Moreover, we make use of the transition probability derived from existing trajectories to generate the possible movements of devices with randomized MAC addresses. The deployment and evaluation in the Asian Civilisations Museum demonstrate the feasibility and utility of the proposed system.

The first two pieces of works make use of the non-invasive technique such as passive scanning to obtain user context information. In the third piece of work, we look at techniques that require users to actively run an APP on their devices. WiFi fingerprint-based indoor localization running on mobile devices has attracted significant attention due to its ease of deployment and low requirement on infrastructure. While there have been lots of solutions proposed by different researchers, there is a lack of effective tools that can easily evaluate different algorithms under a variety of settings.

In our third work, we study the impact of human activity on WiFi Receive Signal Strength (RSS) and factors that impact fingerprint-based localization accuracy. We highlight and present measurements on two phenomena, namely signal blocking effect from the human body and human movement in the environment that can result in significant degradation of localization performance. We propose EvaLoc, a WiFi fingerprint-based localization evaluation tool that helps researchers quantify the localization accuracy degradation under different conditions. Our evaluation in 2 different sites and user feedback from 14 locations show that EvaLoc is able to estimate the localization error more accurately by taking into account factors that impact the quality of the WiFi fingerprints.

Either passive detection or active localization, we make use of WiFi-based information to reason about the behavior and movement of the human. Getting such information provides a new way to learn, understand and improve the human life and enable further application and research on human activity.