Page 65 - Undergraduate Research Opportunities Program Proceedings 2017-18

School of Engineering
Department of Computer Science & Engineering

Indoor Localization and Mobile Computing

Supervisor: CHAN Gary Shueng Han / CSE
Student: LEE Wai Lam / COMP Course: UROP1000, Summer
ZHANG Yichen / CPGBM UROP1100, Summer

With the ability to localize the position of a mobile phone user, a wider range of service can be provided by computer
technology. For example, it can track the path of the user and navigate the user to the destination. Due to the success of
the development of GPS, the localization of outdoor mobile phones is fairly precise. However, it is difficult for the GPS
satellite to access indoor mobile phones, so GPS cannot localize indoor mobile phones accurately. Mobile phones are
used not only in outdoor environment but also indoor situation. Thus, there is a need to develop a system to localize
indoor cellular device.

Supervisor: CHAN Gary Shueng Han / CSE
Student: OH Minseok / SENG Course: UROP1000, Summer

Over the last few decades, Global Navigation Satellite System has enjoyed a tremendous success in multiple
technological domains, mainly revolving outdoor localization. However, due to proliferation of ‘urban canyon
environment’ and relative power inefficiency of GPS scanning, such method is not ideal for indoor localization situations.
This project aims to address this problem, using Wi-Fi positioning system, a more energy efficient and accurate
methodology. The application developed for the project, manually records currently available Access Points’ BSSID and
RSSI, which accumulates and is sorted according to pre-specified criteria. Based on this, geolocational coordinates are
approximated using weighted non-linear least squares optimization algorithm, which provide base dataset to ultimately
locate users solely with their Wi-Fi AP list.

Supervisor: CHAN Gary Shueng Han / CSE
Student: OU Jiefu / SENG Course: UROP1000, Summer

To match the time differences between two speech audio streams (WAV format) - same source but one recorded by
users’ mobile phone, with noise and random time delay than the other. First, I learned and utilized the Mel-frequency
cepstral coefficients (MFCCs) to extract a 13-dimensional vector for each frame, to characterize the speaker’s voice
feature (C++). Second, I came up with a linear comparison algorithm to find the position in the default stream at which
the first 1/10 of the record stream reached its highest similarity compare with the default stream (C++). Finally, I utilized
NDK and JNI in android to combined programs above together as a record-detect delay application.

Supervisor: CHAN Gary Shueng Han / CSE
Student: PANG Hong Wing/ COMP Course: UROP1100, Fall

This project trys to implement a stepcounting algorithm using HMM (Hidden Markov Model), that counts the no. of
steps a person has taken while holding a phone. The no. of steps is used to estimate the distance the phone user has
traveled, which is to be ultmately used in an indoor localization app. The algorithm only takes in the accleration of the
phone, in particular the up-down oscillation as the user walks, into account, and is able to identify if the user is actually
walking or unknowingly shaking his/her phone, which could be mistaken as walking. This report covers the HMM model
behind the algorithm, other steps required for the calculation of probabilities and related experimentation conducted.

55