The Research And Realization Of The G.729 Speech Coding Based On DSP

G.729 speech code is passed in 1996 by ITU-T based on CELP modes which use Conjugate Structure Algebraic Code Excited Linear Prediction algorithm technology is one of the most efficiency compression algorithm of coding rate and synthetic speech quality. It can be widely used in the field of individual mobile communication, multimedia communication, IP phone, satellite communication, ISDN and so on. But it is difficult to realize in real time because of the high algorithmic complexity. With the rapidly developing of DSP technology in recent decades, the computational capability of the chip is gradually improve, and the chip cost is reduced, so the DSP has become the main tools in the field of speech signal processing gradually, and make it become possible to realize the more complex codec algorithm in a single chip. At the same time the algorithm can be easily updated, thus makes it can be update the function without replace the hardware. This paper introduced the algorithm structure of G.729, including part of encode and decode, It put emphasis on some key technologies of G.729 algorithm such as Linear-Prediction, Perceptual Weighting Vector Quantization, Adaptive-codebook Search and Fixed-codebook Search. And describe the structure and the work theory of the TMS320VC5402 DSP. On this basis the system of hardware is presented, focused on the optimization method of the code, use the C source code provided by ITU-T, optimizes the larger operational quantity modules of the source code.Described how to make the hardware and software system work as a whole to complete the real-time test. The test about the optimized code shows that, the result is correct, the synthesized speech has better intelligibility,computing delay is smaller than that the required minimum delay time. It can be implementation on TMS320VC5402 in real-time.This provides a good help for the application of G.729 on low price, low power consumption DSPs.