Speech Coding

On these web pages I try to give an idea of the principles involved in speech coding, and details of commonly used coders. Also links are given to other related pages, and the source code of some common speech codecs.

I have concentrated in these pages on narrowband speech codecs. Such codecs are used to give an efficient digital representation of telephone bandwidth speech. Often the speech is bandlimited to between 200 and 3400 Hz, and is sampled at 8 kHz. An ideal speech codec will represent this speech with as few bits as possible, while producing reconstructed speech which sounds identical, or almost identical, to the uncoded speech. Of course in practice there is always a trade-off between the bit rate of the codec and the quality of its reconstructed speech.

This document is split into the sections shown below. You can either read it in the order suggested below by clicking on the arrows at the bottom of each page, or jump straight to a particular section you are interested in.

The financial support of our current sponsors is gratefully acknowledged. The Engineering and Physical Sciences Research Council (EPSRC), UK, supports our research in the broad field of Wireless Multimedia Communications. Motorola European Cellular Infrastucture Division (ECID), UK and the European Community provide funding in order to conduct research towards third generation Universal Mobile Telecommunications Systems (UMTS) in the framework of the Flexible Integrated Radio System Technology ( FIRST ) project within the ACTS programme.

Back To Home Page On to First Page

Jason Woodard ,
Department of Electronics & Computer Science,
University of Southampton ,
Highfield, Southampton SO17 1BJ