G.729Speech Co-decoding Algorithm Research And Implementation Based On DSP

At present，with the rapid development of modern communication technologies and theincreasing of communication users, high bandwidth efficiency and high communicationquality have become the two main goals pursued by modern communications field.Low cost with higher communication quality and low bit rate speech coder is a tool ormeans for people to achieve both objectives. The Characteristics of speech coding isthat gain the best reconstruction speech signal, and as small as possible codec delay andcompromise computing complexity at a certain coding rate. To meet this need of peopleto communicate, ITU (International Telecommunications Union, ITU) released aprotocl of G.729(or G.729standard) in1996, this protocol adopt a technology based onthe Conjugate Structure Algebraic Code Excited Linear Prediction (CS-ACELP),aSpeech coding standard with a8kbit/s bit rate, a small delay and high quality speechreconstruction. For the moment, the agreement has very broad application prospects,has been involved in the field of personal mobile communications, multimedia digitalnetwork communications, telephone, mobile satellite communications, integratedservices digital communication (ISDN) and so on. At the same time, the agreement isone of the speech codec standards that ITU recommended strongly to the globaltelecommunications sector, but because of its higher algorithm complexity, real-time isstill the main application obstacle.The main subject of this study is based on the DSP platform to realize real-timeencoding and decoding algorithm. Digital signal processor (DSP) has a high-speedcomputing capabilities and a wide range of applications in the field of informationprocessing. In consideration of the real-time, data size and precision as well as cost, thisstudy adopt a processing chip of high-performance fixed-pointDSP----TMS320VC5402,which producted by the Texas Instruments Incorporated (TI).This topic First introduced the technology related to the encoding and decodingalgorithm principle of G.729, then the algorithm itself, namely that the ConjugateStructure----Algebraic Code Excited Linear Prediction algorithm by means of sourcecode to specificly analysis its implementation process. Then analysis the source code ofG.729standard that the International Telecommunication Union released at the sametime, and serialize operation to source code before transplantation to reach thecompression ratio of16:1. Followed by detailing the codec hardware platform,introducing the interface and configuration between the voice capture chip and the mainprocessing chip, and then introduced the software design of the hardware platformwhich has been set up. Finally, according to the characteristics of DSP, combined withthe characteristics of encoding and decoding algorithm, proposed a series ofoptimization methods which optimized mainly in four aspects: the compiler-leveloptimization, algorithm-level optimization, the C language level optimization and assembly-level optimization, and gave its results achieved in real time in DSP platformafter optimization.