Date of Award

2019

Publication Type

Doctoral Thesis

Degree Name

Ph.D.

Department

Electrical and Computer Engineering

First Advisor

Esam Abdel-Raheem

Keywords

Adaptive filters, Blind equalization, Lattice structure, Least mean squares, Recursive least squares, Variable tap-length

Rights

info:eu-repo/semantics/embargoedAccess

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

Today the usage of digital signal processors has increased, where adaptive filter algorithms are now routinely employed in mostly all contemporary devices such as mobile phones, camcorders, digital cameras, and medical monitoring equipment, to name few. The filter tap-length, or the number of taps, is a significant structural parameter of adaptive filters that can influences both the complexity and steady-state performance characteristics of the filter. Traditional implementation of adaptive filtering algorithms presume some fixed filter-length and focus on estimating variable filter's tap-weights parameters according to some pre-determined cost function. Although this approach can be adequate in some applications, it is not the case in more complicated ones as it does not answer the question of filter size (tap-length). This problem can be more apparent when the application involves a change in impulse response, making it hard for the adaptive filter algorithm to achieve best potential performance. A cost-effective approach is to come up with variable tap-length filtering scheme that can search for the optimal length while the filter is adapting its coefficients. In direct form structure filtering, commonly known as a transversal adaptive filter, several schemes were used to estimate the optimum tap-length. Among existing algorithms, pseudo fractional tap-length (FT) algorithm, is of particular interest because of its fast convergence rate and small steady-state error. Lattice structured adaptive filters, on the other hand, have attracted attention recently due to a number of desirable properties. The aim of this research is to develop efficient adaptive filter algorithms that fill the gap where optimal filter structures were not proposed by incorporating the concept of pseudo fractional tap-length (FT) in adaptive filtering algorithms. The contribution of this research include the development of variable length adaptive filter scheme and hence optimal filter structure for the following applications: (1) lattice prediction; (2) Least-Mean-Squares (LMS) lattice system identification; (3) Recursive Least-Squares (RLS) lattice system identification; (4) Constant Modulus Algorithm (CMA) blind equalization. To demonstrate the capability of proposed algorithms, simulations examples are implemented in different experimental conditions, where the results showed noticeable improvement in the context of mean square Error (MSE), as well as in the context of convergence rate of the proposed algorithms with their counterparts adaptive filter algorithms. Simulation results have also proven that with affordable extra computational complexity, an optimization for both of the adaptive filter coefficients and the filter tap-length can be attained.

Available for download on Sunday, April 05, 2020

Share

COinS