Optimization Of G.729 Annex B Speech Coder And Its DSP Implementation

Speech of coding is a very important technology in digital speech communication and low-bit-rate speech codec plays an important role in resolving limit of bandwidth resource. Coding of speech at 8kbit/s algorithm G.729 was standardized by the International Telecommunication Union-Telecommunication Standardization Sector, based on conjugate-structure algebraic-code-excited linear-prediction (CS-ACELP). G.729 is one of the most complicated algorithms of speech compression, which uses kinds of advanced technologies in speech coding and compression. Annex B to ITU-T recommendation G.729 defines the Voice Activity Detection(VAD), Discontinuous Transmission(DTX) and Comfort Noise Generator(CNG) algorithms. These algorithms are used during silence periods of speech to reduce ratio of transmission and improve ratio of use in signal channel. It's a best silence compression scheme for G.729.The paper studies and analyzes detailedly the theory of G.729 speech codec and key technologies of G.729 Annex B. On the base of being familiar with C language standard source code of G.729 Annex B, the performance of its code is evaluated by the profiler which is a tool of analysing code. Code optimization is executed for some functions waste a lot of time in processing by way of C languange level and assembly level. Under the condition of guaranteeing favorable speech quality, it is necessary to reduce the code quantity of recommendation G.729B. After the algorithm is optimized, the amended code is transplanted to the platform of DSP TMS320VC5416. Finally the algorithm of G.729 Annex B is implemented in real time on the fixed DSP. At last we test the system and draw some conclusions. The results of expriment show that the decoded speech signal whose definition and nature characteristic are all good by the method of subjective audition, and its reconstructed wave shape and frequency spectum are similar to the original ones.