Date of Award

6-2012

Degree Type

Thesis

Degree Name

Master of Science (MS)

Department

Electrical Engineering and Computer Science

Advisor(s)

Pramod K. Varshney

Keywords

Decision Fusion, Hypotheses Testing, Modulation Classification, Parameter Estimation, Wireless Communication Networks, Wireless Sensor Networks

Subject Categories

Electrical and Computer Engineering

Abstract

Modulation classification is a crucial step between data recognition and demodulation in Cognitive Radio systems. In this thesis, automatic blind modulation classification approaches designed for multiple modulation classification in distributed sensor network are discussed. Optimal Bayesian Approach and Likelihood Based Detection are the main mathematical foundations. First, we build a channel and signal model under the assumption of Additive White Gaussian Noise (AWGN) and Rayleigh Fading Channel. For the distributed detection scheme, we compare the performance of Chair-Varshney Fusion Rule with Majority Fusion Rule in a coherent communication environment. For a more general scenario with embedded unknown channel parameters, Hybrid Likelihood Ratio Test method with different types of estimations are studied. Treating transmitted symbols as hidden variables, we propose an Expectation Maximization Algorithm based approach to determine maximum likelihood estimates (MLE) so that we can have closed form expressions of MLEs. Centralized data fusion based classifier equipped with EM algorithm based maximum likelihood estimates is then evaluated via comparison to the classifier equipped with Method of Moments (MoM) estimates. From computer simulation experiments, we conclude that EM algorithm based MLE can efficiently obtain better classification performance than Moments-Based estimation especially in low Signal-to-Noise Ratio environments.

Access

Open Access