Equalization Using Graphical Models

We examine the use of graphical models for the equalization of digital communication channels with memory. Graphical models provide a framework in which the structure of large systems can be exploited to derive efficient estimation algorithms. Furthermore, properties of a graph on which an algorithm is based can be used for analysis. We use factor graphs to develop efficient implementations of various equalization algorithms, including an unconstrained and a constrained linear minimum mean squared error equalizer and a generalized decision feedback equalizer. In addition to providing practical algorithms, the factor graph framework yields insight into their mechanics and interrelationships. We then consider algorithms for turbo equalization, using both graphical models and an algebraic approach to derive appropriate equalizers. In all cases, associated graphs are used for performance analysis. Finally, we consider universal piecewise linear equalization, adapting the well-known context-tree weighting algorithm. In addition, we generalize this approach by introducing context graphs that allow for a trade-off between modeling power and computational complexity. Although we focus primarily on equalization algorithms, many of the approaches considered are more general and can be applied to a wide variety of estimation problems.

Name thesis:degree_name

Ph.D.

Level thesis:degree_level

Dissertation

Discipline thesis:degree_discipline

Electrical and Computer Engineering

Grantor

University of Illinois at Urbana-Champaign

Year dc:date

2015

Author dc:creator

• Drost, Robert James

Contributors dc:contributor

• Singer, Andrew C.

• Engineering, Electronics and Electrical

Language dc:language

eng

Identifier

(MiAaPQ)AAI3290223

OAI identifier oai:identifier

oai:www.ideals.illinois.edu:2142/81031