Date of Award

12-1991

Document Type

Thesis

Degree Name

Master of Science

Department

Department of Electrical and Computer Engineering

First Advisor

Rob Williams, PhD

Abstract

Two Walsh-domain dyadic convolution adaptive filters are developed using a circular convolution frequency-domain filter (FDF1) and the Fast LMS adaptive filter (FDF2): WDF1 and WDF2 respectively. General theory of time- domain adaptive filters and a theoretical analysis of the FDF1, FDF2, WDF1, and WDF2 filters are presented. WDF1 and WDF2 software implementation are shown to be error free. A time-domain filter (TDF) and a FDF2 frequency-domain filter (FDF) are implemented for comparison testing. The WDF1, WDF2, TDF, and FDF filters are tested using time-shifted sinusoidal and rectangular noisy and noiseless signals. WDF1 and WDF2 are shown to converge faster and produce less error filtering discontinuous signals, relative to the TDF and FDF performance. WDF1 and WDF2 are shown to converge slower and produce more error filtering continuous signals, relative to TDF and FDF performance. WDF1 is shown to perform better for noiseless signals, relative to WDF2 performance. WDF2 is shown to perform better for noise signals, relative to WDF1 performance. WDF1 and WDF2 filtering performance was shown to degrade with increasing time shift. A processing speed comparison showed WDF1 to be faster than the TDF, FDF, and WDF2 filters.

AFIT Designator

AFIT-GE-ENG-91D-15

DTIC Accession Number

ADA243701

Comments

The author's Vita page is omitted.

Download

Share

COinS