Please use this identifier to cite or link to this item:
Title: Some studies in data compression using wavelet transform with special reference to Indian language speech
Researcher: Vithalani, C H
Guide(s): Dave, Himanshu B
Keywords: Electronics Engineering
Indian language speech
wavelet transform
data compression
Human auditory system
Upload Date: 21-Aug-2012
University: Gujarat University
Completed Date: 2004
Abstract: The focus of this thesis is to compress the digital speech using wavelet transform. The various digital speech compression algorithms using wavelet transforms are implemented. The wavelet transform is a valuable tool for many engineering applications. We have selected wavelet analysis for the speech compression, as it is useful from a psychoacoustic point of view. It has close relation to many logarithmically structured perceptual scales such as pitch and loudness. The constant Q properties of the wavelets mirrors good perceptual qualities. The wavelet transform is useful to remove redundancies and irrelevancies present in the speech signal for the compact representation. There are two popular methods of wavelet transform. One is classical sub-band coding method also known as filter bank method in which signal is decomposed into basis of wavelet functions using high pass and low pass filters. Output of high pass filter represents detail components and output of low pass filter represents approximate components. Other method is lifting scheme of the wavelet transform. The Lifting scheme also decompose speech signal into approximate and details components. In order to compress the speech signal, small wavelets are removed by thresholding i.e. the details below some threshold value are replaced by zero. We have worked out details of integer wavelet transform based on lifting scheme of wavelet transform. The lifting scheme is fast compared to filter bank method. We have implemented and tested various wavelet transform algorithms for the speech compression using filter bank method and lifting scheme for different wavelets like Haar, Daubechies series (Daub-4, Daub-6, Daub-8, Daub-12, Daub-16) and Cohen-Daubechies-Feauveau bi-orthogonal wavelet with English, Hindi, Gujarati and Sanskrit language words and sentences. Speech quality is measured for different wavelets using subjective and statistical method.
Pagination: 239p.
Appears in Departments:Department of Engineering

Files in This Item:
File Description SizeFormat 
01_title.pdfAttached File36.04 kBAdobe PDFView/Open
02_acknowledgement.pdf11.95 kBAdobe PDFView/Open
03_abstract.pdf16.59 kBAdobe PDFView/Open
04_contents.pdf19.26 kBAdobe PDFView/Open
05_chapter 1.pdf46.34 kBAdobe PDFView/Open
06_chapter 2.pdf727.15 kBAdobe PDFView/Open
07_chapter 3.pdf271.89 kBAdobe PDFView/Open
08_chapter 4.pdf310.01 kBAdobe PDFView/Open
09_chapter 5.pdf255.13 kBAdobe PDFView/Open
10_chapter 6.pdf457.81 kBAdobe PDFView/Open
11_chapter 7.pdf37.81 kBAdobe PDFView/Open
12_appendix.pdf456.19 kBAdobe PDFView/Open

Items in Shodhganga are protected by copyright, with all rights reserved, unless otherwise indicated.

Altmetric Badge: