Indoor crowdsourced Wi-Fi fingerprinting with network embedding

The Hong Kong University of Science and Technology
Department of Computer Science and Engineering

Final Year Thesis Oral Defense

Title: "Indoor crowdsourced Wi-Fi fingerprinting with network embedding"

by

ZHAO Ziqi

Abstract:

WiFi fingerprints are crucial in prevalent smart city applications, such 
as indoor localization and WiFi monitoring. Recent works on fingerprinting 
with crowdsourced data usually leverage inertial sensors, or radio 
propagation models to label collected signals. They may suffer from 
accumulative error of sensors, non-line-of-sight environments where models 
would fail. Other works use manifold alignment to estimate locations with 
sparse fingerprints, but may not be easily scalable due to the computation 
complexity. In this project, we propose a scalable crowdsourced 
fingerprinting system based on pure WiFi signals without any assumption on 
signal propagation models. The system first uses network embedding to 
infer dimension-reduced representations of WiFi signals, and then matches 
these learned representations onto the map with sporadic location labels. 
Extensive experiments are conducted and various evaluation criteria are 
applied to show the performance of the proposed method.


Date            : 20 May 2020 (Wednesday)

Time            : 14:00 - 14:40

Zoom Meeting    : https://hkust.zoom.us/j/947854476

Advisor         : Prof. CHAN Shueng-Han Gary

2nd Reader      : Dr. WONG Raymond Chi-Wing