Date of Award

2015

Publication Type

Master Thesis

Degree Name

M.A.Sc.

Department

Electrical and Computer Engineering

First Advisor

Rashid Rashidzadeh

Second Advisor

Roberto Muscedere

Keywords

Indoor positioning, Paritcle filter, Support vector machine, WLAN

Rights

info:eu-repo/semantics/openAccess

Creative Commons License

Creative Commons Attribution-Noncommercial-No Derivative Works 4.0 License
This work is licensed under a Creative Commons Attribution-Noncommercial-No Derivative Works 4.0 License.

Abstract

In this thesis, an indoor localization method using on-line independent support vector machine (OISVM) classification method and under-sampling techniques is proposed. The proposed positioning method is based on the received signal strength indicator (RSSI) of Wi-Fi signals. A new under-sampling algorithm is developed to address the imbalanced data problem associated with the OISVM, and a kernel function parameter selection algorithm is introduced for the training process. The time complexity of both the training process and the prediction process are decreased. Comparative experimental results indicate that the training speed and the prediction speed are improved by at least 10 times and 5 times, respectively. Furthermore, through on-line learning, the estimation error is decreased by 0.8m. Such an improvement makes the proposed method an ideal indoor positioning solution for portable devices where the processing power and the memory capacity are limited. A new Particle Filter (PF) scheme for indoor localization using Wi-Fi received signal strength indicator (RSSI) and inertial sensor measurements has also been presented. RSSI is affected significantly by multipath fading, building structure and obstacles in indoor environments. The information provided by inertial sensors combined with the proposed particle filter are used to develop a positioning algorithm supporting a smooth and stable localization experience. To differentiate similar fingerprints, a single-hidden layer feedforward networks (SLFNs) is used to model the multiple probabilistic estimations and to improve the performance of the PF. A new initialization algorithm using Random Sample Consensus (RANSAC) has also been presented to reduce the convergence time. Experimental measurements were carried out to determine the performance of the proposed algorithm. The results indicate that the positioning error falls to less than 1.2 (m).

Share

COinS