Multimodal Learning and Single Source WiFi Based Indoor Localization

Date of Award


Degree Name

M.S. in Computer Engineering


Department of Electrical and Computer Engineering


Advisor: Feng Ye


With the rapid development of high speed Internet and Internet of Things (IoT) ap-plications, the demand of indoor localization technology is increasing over years. Well-developed indoor localization technologies can bring significant changes to industries suchas health-care, manufacturing, and security, etc. Wi-Fi fingerprint-based indoor localizationis becoming more and more popluar thanks to pervasive deployment of Wi-Fi access pointsand low maintenance cost. However, Wi-Fi fingerprint-based method requires a databaseof collected signal information from all interest points before hand, which means that theprocess of data preparation is time-consuming and labor-intensive. Meanwhile, due to thedynamic nature of environment, the persistence of localization system is unstable, thusfrequent data re-acquistion and model re-modeling are needed. In addition, current Wi-Fi fingerprint-based methods require multiple WiFi sources, which leads to the increasingamount of cost when constructing the localization system. Therefore, to tackle these is-sues, a multimodal learning and single source Wi-Fi based indoor localization system isproposed. The proposed system contains three components: Firstly, a moving object de-tection approach is applied for video processing to generate location labels. Secondly, asingle-source Wi-Fi based localization model is developed using the collected signal dataas well as the autonomously generated location labels. Lastly, a path tracking scheme isproposed to demostrate efficacy of the proposed localization model. Computer based simu-lation results show that the proposed system provides effective solutions to current indoorlocalization problems.


Computer Engineering, Electrical Engineering, deep learning, indoor localization, multimodal learning

Rights Statement

Copyright © 2020, author