Abstract

This thesis is devoted to the investigation of effective implementations of a modified version of a well-established fixed-point data-dependent VAD-CNG algorithm of Nortel Networks. In this study, the algorithm to be implemented is first described in detail. Then the target platform, a TMS320C5402DSK DSP board, along with its real-time operating system (DSP/BIOS) and its associated analysis tools, is introduced. Two implementation schemes are presented in this thesis. The first one is a direct implementation of the modified algorithm on the DSP board. The second one is an implementation, wherein some optimizations that target the reduction of the implementational complexity of the algorithm are introduced. Several DSP/BIOS real-time analysis tools provided by the DSP board and some speech samples are used to test the performance of the implementations. Experimental results are presented to show the effectiveness of the optimizations for the modified VAD-CNG algorithm. These results show that over 80% of the reduction in the implementational complexity is achieved through the proposed optimizations, making it possible to incorporate such a VAD-CNG algorithm into a practical real-time voice communication system. A real-time audio codec system is built in the laboratory to demonstrate the real-time implementation of this algorithm. This demonstration system should also serve as a useful experimental tool for further investigation on this topic. (Abstract shortened by UMI.)