Design Of Embedded Speech Capture And Compression System Based On G.726/G.729

In modern communication systems, the frequency resources are very valuable, in order tosave transmission bandwidth, high-quality and low-rate speech compression technology becomesparticularly important. ITU-T is an international standardization organization of communicationsindustry. As a branch of this organization, speech-coding technology has established a great lotof low-rate speech compression standards, including G.721, G.723, G.726, G.728, G.729 and etc.which G.726 24kbit/s speech compression standard use the ADPCM algorithm, so with the lowcomputational complexity, but the high coding rate, low speech quality. G.729 speechcompression standard use the CS-ACELP algorithm which based on CELP. provide low bit-rate,high-quality, low-coding rate, but relative to the G.726 24kbit/s encoding, the computationalcomplexity is much higher.With the 32-bit embedded processor chip in a substantial increase in performance and pricedecline, embedded systems are more and more widely used. In this system, using the GM8180chip, which developed by Grain Media Ltd. the chip with a highly integrated Faraday FA626main chip, ARM 32-bit RISC, frequency of up to 500M. 32kB instruction cache and 32kB datacache. This enabled it to complete a variety of speech codecs such as AMR, AAC, G.726, G.729and so on.This article discusses the performance of a variety of low-rate speech coding standards, inorder to maximize the savings in system resources, Using two kinds of encoding G.726/G.729alternative coding scheme (Requiring high quality of the speech came time to select G.729,otherwise, select the G.726 encoding method). And then focuses on the G.726/G.729 codecprinciple, On this basis, put forward two kinds of optimization algorithms: coarsening of theopen-loop pitch search, optimization of the LSP coefficients quantization algorithm, effective inreducing the amount of arithmetic operations, and then build MATLAB simulation platform tovalidate the feasibility of the algorithm.Second, the introduction of the embedded processor chipGM8180 and embedded processing systems Linux-2.6.14 performance and features, and buildthe system hardware and software platform, and then complete the transplantation of the G.726/G.729 source code provided by ITU organizations to the platform. and optimizes theoriginal code according to the study and application of optimizing strategy. And last, the result ofthe coding efficiency, speech quality testing and evaluation show that the optimizer complete thespeech codec function well, coding efficiency and speech quality has been achieved the desiredobjective.