TY  - THES
T1  - Development of a parametric speech model for wideband speech coding
A1  - Tan, Emerson Ching
LA  - English
YR  - 2003
UL  - https://tuklas.up.edu.ph/Record/UP-99796217607612171
AB  - Wideband speech offers a more natural sounding voice in any telecommunication system. Typical applications include teleconferencing and videoconferencing. Wideband speech covers 50 Hz to 7000 Hz of the spectrum and is sampled at 1g kHz. Various methods for coding wideband speech have been proposed and some have been accepted for standardization. Most of the proposed wideband speech coders are either waveform coders or hybrid coders. Parametric coders offer the lowest bit rates with acceptable speech quality. A wideband speech model is an integral part of a wideband parametric coder. In this thesis, wideband speech models are developed by extending popular narrowband speech models through the observation of peculiar characteristics in wideband speech. These popular narrowband speech models are the linear predictive model and the sinusoidal model. Several configurations of wideband linear predictive model and wideband sinusoidal model are implemented. Through subjective testing for quality and intelligibility, six configurations are found to be acceptable. These wideband speech models can be used in future research for the development of a very low bit rate wideband speech coder.
CN  - LG 995 2003 E64 T36
KW  - Speech processing systems.
KW  - Speech synthesis.
KW  - Coding theory.
ER  -