The New Jersey Institute of Technology's
Electronic Theses & Dissertations Project

Title:	Sparse representation for audio noise removal using zero-zone quantizers
Author:	Mittal, Neha
View Online:	njit-etd2010-007 (xi, 52 pages ~ 3.0 MB pdf)
Department:	Department of Electrical and Computer Engineering
Degree:	Master of Science
Program:	Electrical Engineering
Document Type:	Thesis
Advisory Committee:	Akansu, Ali N. (Committee co-chair) Haddad, Richard A. (Committee member) Sinha, Deepen (Committee co-chair)
Date:	2010-01
Keywords:	Noise removal Zero zone quantization
Availability:	Unrestricted
Abstract:	In zero zone quantization, bins around zero are quantized to a zero value. This kind of quantization can be applied on orthogonal transforms to remove the unwanted or redundant signal. Transforms reveal structures and properties of a signal and hence careful application of a zero zone over the transform coefficients leads to noise removal. In this thesis, such quantizers are applied over Discrete Fourier Transform and Karhunen Loeve Transform coefficients separately, and outputs compared. Further, the localization of the zero zones to certain frequencies leads to better performance in terms of noise removal. PEAQ (Perceptual Evaluation of Audio Quality) scores have been used to measure the objective quality of the denoised signal.