Research And Implementation On G.729AB Speech Codec System

In modern voice communication systems, transmission efficiency and voice quality are the research hot spots. Low-rate and high quality speech vocoder is in demand. To meet these requirements, in 1996, ITU-T release G.729 recommendation, coding of speech at 8kbit/s using Conjugate Structure Algebraic Code Excited Linear Prediction (CS-ACELP) algorithm. G.729 Annex A is the reduced-complexity version of the G.729 speech codec. G.729 Annex B is a silence compression scheme for G.729. G.729A is adapted to operate with G.729B, which reduces the average bit-rate with no significant effect to voice quality. G.729AB achieves the toll quality, which is equivalent to that of 32kbits/s ADPCM for most operating conditions.In the base of research on the algorithm of G.729AB, this thesis proposes realtime implementation of G.729AB speech codec system based on StarCore SC140e DSP core of Freescale Semiconductor. In the first chapter, the developing actuality and tendency of speech codec technology is introduced. Then, the key technology of Code Excited Linear Prediction (CELP) speech codec is analyzed, including Linear Prediction, Vector Quazition (VQ) and Perceuptual enhancement. The algorithms of G.729 along with G.729A and G.729B are also analyzed in briefly. By grasping StarCore DSP core features and SC140e assembly language, the software implementation of G.729AB is analyzed in detail, including the porting strategy and optimization skills of speech codec system on fix-point DSP. The system is optimized in both C and assembly stage.Finally, the test plan is designed according to the project requirements and system test is performed. The optimized speech codec system passes all ITU-T standard test vectors, achieveing realtime systems performance and meeting with all the functional requirements.