Variable Bit Rate Speech Coding

Abstract

During the last several years, there has been a dramatic growth of digital services, such as digital wireless and wireline communications, satellite communications, multimedia applications, voice over internet protocol (IP) and digital voice storage. Such systems require special techniques with high to average quality speech and average to low bit rate to make an efficient coding of speech signal before transmission or storage. The majority of these coders employ algorithms that are based on Code Excited Linear Prediction (CELP) coder.

This thesis presents a new Rate Determination Algorithm for a Selectable Mode Vocoder . This Algorithm is based on the new Third Generation Partnership Project 2 (3GPP2) Study Group 16 (SG16) proposal to implement a Selectable Mode Vocoder.

A new Rate Determination Algorithm (RDA) has been developed in order to reduce the bit rate needed to quantize the spectral parameters of the existing coders. This technique uses a combination between Voice Activity Detection based on voiced-mivoiced classification of speech frames and Rate Determination Mechanism. Also a new technique has been introduced in the calculation of the voice activity detection flag to allow a more robust voiced-unvoiced classification. This method is based on enhancing the voice activity detection by using sliding windows with size equivalent to the speech frame size under process. The Rate Determination