Analysis And Optimization Of ITU-T G.723.1 Speech Coder

In 21 century, communication should supply all kinds of ways to satisfy people's need in exchanging information availably. As speech is the most important, efficient and convenient way in communications, speech communication holds the importance in modern digital communication system. The frequency source has become more and more valuable with the development of information society, so more and more speech compress coding technologies come into being to decrease bit rate in transmission, and play an important role in mobile communication, voice over IP, multimedia information transmission, secrecy communication, etc. Many international standards of such speech codec have been set up, and the low bit rate speech coding and its applications have become the main field for research.This paper focuses on the principles of ITU-T G.723.1 dual rate speech codec standard, with analyzing and algorithm design of the main modules. Basing on Analysis-by-Synthesis, implemented the speech compressing and decoding by Algebraic Code Excited Linear Prediction (ACELP) and Multi-Pulse Maximum Likelihood Quantization principles, and proved the algorithm to be feasible.Base on that, the paper puts forward to optimize the modules with large computational complexity, such as pitch estimation module, the adaptive and fixed code-book research modules, and describes in detail the software design as well as some key techniques in programming optimization. Using the test vectors from the recommendation G.723.1, the optimized algorithm has been tested and analyzed.In the end, the paper presents the speech testing of the optimized algorithm to meet the real-time need of the speech communication. The testing results show the optimized algorithm can represent the speech with a high quality using a limited amount of complexity. The decoding signals show a good Mean Opinion Score and robustness, and proved the codec to be practical applicable.