Algorithms for Blind Equalization Based on Relative Gradient and Toeplitz Constraints

Loading...
Thumbnail Image
Degree type
Doctor of Philosophy (PhD)
Graduate group
Electrical & Systems Engineering
Discipline
Subject
Adaptive filter
Blind equalization
Blind source separation
Independent component analysis
Signal processing
Electrical and Electronics
Funder
Grant number
License
Copyright date
2016-11-29T00:00:00-08:00
Distributor
Related resources
Contributor
Abstract

Blind Equalization (BE) refers to the problem of recovering the source symbol sequence from a signal received through a channel in the presence of additive noise and channel distortion, when the channel response is unknown and a training sequence is not accessible. To achieve BE, statistical or constellation properties of the source symbols are exploited. In BE algorithms, two main concerns are convergence speed and computational complexity. In this dissertation, we explore the application of relative gradient for equalizer adaptation with a structure constraint on the equalizer matrix, for fast convergence without excessive computational complexity. We model blind equalization with symbol-rate sampling as a blind source separation (BSS) problem and study two single-carrier transmission schemes, specifically block transmission with guard intervals and continuous transmission. Under either scheme, blind equalization can be achieved using independent component analysis (ICA) algorithms with a Toeplitz or circulant constraint on the structure of the separating matrix. We also develop relative gradient versions of the widely used Bussgang-type algorithms. Processing the equalizer outputs in sliding blocks, we are able to use the relative gradient for adaptation of the Toeplitz constrained equalizer matrix. The use of relative gradient makes the Bussgang condition appear explicitly in the matrix adaptation and speeds up convergence. For the ICA-based and Bussgang-type algorithms with relative gradient and matrix structure constraints, we simplify the matrix adaptations to obtain equivalent equalizer vector adaptations for reduced computational cost. Efficient implementations with fast Fourier transform, and approximation schemes for the cross-correlation terms used in the adaptation, are shown to further reduce computational cost. We also consider the use of a relative gradient algorithm for channel shortening in orthogonal frequency division multiplexing (OFDM) systems. The redundancy of the cyclic prefix symbols is used to shorten a channel with a long impulse response. We show interesting preliminary results for a shortening algorithm based on relative gradient.

Advisor
Saleem A. Kassam
Date of degree
2016-01-01
Date Range for Data Collection (Start Date)
Date Range for Data Collection (End Date)
Digital Object Identifier
Series name and number
Volume number
Issue number
Publisher
Publisher DOI
Journal Issue
Comments
Recommended citation