Dissertation/Thesis Abstract

Nonparametric regression using wavelet transformations and noise reduction in audio signals
by Brothers, Jonathan, M.S., California State University, Long Beach, 2009, 173; 1466191
Abstract (Summary)

A major concern in audio processing is the reduction or dampening of white noise. Any technique for noise reduction must maintain the musical integrity of the signal. In audio signal processing, thresholding wavelet coefficients is a well established technique for noise reduction. This technique is also an established statistical procedure for generating a nonparametric regression. The primary objective of my work is to explore the use of statistical based procedures in audio processing.

Two general threshold rules and four threshold selection procedures are defined and explored. The bias, variance, and risk of the threshold rules are used as points of comparison. These concepts are expanded and used to develop a modeling procedure. Since the model is valid for a large class of signals, the model may be applied in a variety of fields. The particular application considered is the reduction of noise in digital audio signals. Both musical recordings and artificially generated signals are used to demonstrate the model. The final results are compared with a parametric model.

Indexing (document details)
Advisor: Ebneshahrashoob, Morteza
School: California State University, Long Beach
School Location: United States -- California
Source: MAI 47/05M, Masters Abstracts International
Subjects: Mathematics
Publication Number: 1466191
ISBN: 9781109168105
Copyright © 2019 ProQuest LLC. All rights reserved. Terms and Conditions Privacy Policy Cookie Policy