Junjie Chen

Graduation Semester and Year

2014

Language

English

Document Type

Dissertation

Degree Name

Doctor of Philosophy in Electrical Engineering

Department

Electrical Engineering

First Advisor

Qilian Liang

Abstract

Big data requires exceptional technologies to efficiently process large quantities of data within tolerable elapsed times. While the potential benefits of Big Data are significant, there are still a lot of technical challenges that must be solved to fully realize this potential for Big Data.For processing, transporting and storing large data sets of enormous sizes, data need to be greatly compressed. In this thesis, several sparse sensing algorithms - compressive sensing (CS), co-prime sampling and nested sampling are studied for Big Data, in theory and applications.Error performance bounds of noisy compressive sensing are derived based on information theory and estimation theory. Information rate distortion function is a measure as the number of bits per symbol to be stored or transmitted under the constraint of a distortion. Rate distortion performance for scalar quantization of measurement observation is derived. Based on this, reconstruction rate distortion is also studied for CS. In addition, we study the real-world applications of CS in Big Data, to Synthetic Aperture Radar (SAR), radar sensor networks (RSNs), and underwater acoustic sensor networks (UWASNs).Besides, properties of two new sparse sampling schemes, i.e., coprime sampling and nested sampling are investigated, such as rate distortion function, since sparse sampling can cause possible distortion because less number of samples are used. Theoretical analysis of how these two sparse sampling methods affect the power spectral density is given as well. A secure transmission scheme for Big Data based on coprime sampling and nested sampling is provided as well.In addition, a hybrid sparse sampling approach is proposed, which combines nested sampling and compressive sensing to reduce the number of symbols, and rate distortion function is used as a criteria to determine how many bits should be used to represent the symbols during this process. We show that with this hybrid approach, less number of bits is required to represent the sensed information.Finally, since LTE has been a Big Data consumer with ample data, how to allocate resources in the era of Big Data in telecommunications becomes a new issue. A bandwidth allocation method based on smartphone users personality traits and channel condition is studied in a unified mathematical framework in this dissertation.In conclusion, facing "huge storage and bandwidth costs" challenges for Big Data, several approaches of sparse sensing in Big Data are studied, in theory and applications. Summary of contributions in this dissertation and future works are provided at the end.

Disciplines

Electrical and Computer Engineering | Engineering

Comments

Degree granted by The University of Texas at Arlington

Download

Share

COinS