University of Surrey

Test tubes in the lab Research in the ATI Dance Research

Sinusoidal Model Based Low Bit Rate Speech Coding for Communication Systems.

Yeldener, Suat. (1993) Sinusoidal Model Based Low Bit Rate Speech Coding for Communication Systems. Doctoral thesis, University of Surrey (United Kingdom)..

[img]
Preview
Text
11009258.pdf
Available under License Creative Commons Attribution Non-commercial Share Alike.

Download (8MB) | Preview

Abstract

There is an increasing demand for providing speech communications to a person away from his wire-line telephone. Since there is growth in the traffic of communications systems, development of spectrally efficient transmission systems are necessary. To design a spectrally efficient communication system requires minimising the speech transmission bit rate using speech compression (or coding) techniques. There has been considerable interest in the development of low bit rate, high quality speech coding systems. New developments in digital speech communication are evolving at a time when major advances in electronic device technology promises to make implementation practical. This increased capability and decreased cost of digital hardware has prompted an increased interest in more complex and sophisticated coding algorithms which offer better coding quality at lower bit rates. In order to achieve this improved performance, coding techniques must exploit, to a greater degree, information about the mechanisms of speech production and speech perception. Applications for speech coding systems include voice mail, low bit rate digital communications (Digital Cellular Mobile Radio and Portable Communications, Mobile Satellite Communications, Personal Communication Systems and VSAT Networks) and high security telephony such as military communications. For some of these applications, sophisticated algorithms have been developed. Particular classes of these including Analysis-by-Synthesis Linear Predictive Coding (AbS-LPC) and Sinusoidal Model Based Speech Coding, have been subjects of active world-wide research. AbS-LPC such as Code Excited Linear Prediction (CELP) has potential of producing near toll quality speech at bit rates in the range of 6 to 9. 6 kb/s. To encode speech signals below these rates, AbS-LPC schemes are not applicable, due to the large amount of quantisation noise. In this thesis, therefore, Sinusoidal Model Based Speech Coding (SMB-SC) algorithms are investigated to make very low bit rate speech coding possible. The aim of the research is to produce high communication quality speech at 4. 8 kb/s and below by considering aspects of speech analysis, modelling and quantisation. The SMB-SC algorithms operate by exploiting the spectral envelope representation and periodicity of speech signals. All-pole model representation of the speech spectral envelope is examined and various all-pole model derivations are presented. Accurate representation of periodic speech segments is essential for synthesising high quality digital speech at very low bit rates. For this purpose, robust pitch estimation algorithms are investigated that play a fundamental role in SMB-SC algorithms. The popularity of SMB-SC algorithms lies in the fact that they achieve highly periodic speech at low bit rates and are also able to process the aperiodic and transition type of speech signals. The SMB-SC systems include Sine Wave Excited Linear Prediction (SWELP), Multi-Band Excitation (MBE) and Multi-Band Linear Predictive (MB-LPC) speech coders that are capable of producing good quality speech at 4. 8 kb/s and below. All of these coders use sophisticated encoding and decoding methods which will be detailed in this thesis.

Item Type: Thesis (Doctoral)
Divisions : Theses
Authors : Yeldener, Suat.
Date : 1993
Additional Information : Thesis (Ph.D.)--University of Surrey (United Kingdom), 1993.
Depositing User : EPrints Services
Date Deposited : 30 Apr 2019 08:07
Last Modified : 20 Aug 2019 15:31
URI: http://epubs.surrey.ac.uk/id/eprint/851104

Actions (login required)

View Item View Item

Downloads

Downloads per month over past year


Information about this web site

© The University of Surrey, Guildford, Surrey, GU2 7XH, United Kingdom.
+44 (0)1483 300800