The Implementation And Optimization Of G.723.1 Speech Coder Based On Blackfin Platform

ITU-T G.723.1 algorithm is a dual rate speech codec standard. This coder has two kinds of bit rates, i.e., 5.3 Kbps and 6.3 Kbps,which corresponds to two different algorithms. For 5.3 Kbps codec, Algebraic Code Excited LinearPrediction (ACELP) principles are used,while for the 6.3 Kbps codec, MultiPulse-Maximum Likelihood Quantization (MP-MLQ) techniques are used. Although this standard has given the reference code which the G.723.1 algorithm realizes, these codes use the C language to carry on, and has the very good portability, but these codes have not considered the concrete processor's platform characteristic, and the algorithm itself considered merely the clear description and has not considered the algorithm efficiency. Resulting in large delay algorithm,and the efficiency is low, simply can not apply to actual product.This thesis presents a full duplex,real time implementation of ITU-G.723.1 algorithm using the BF 532 DSP chip with the uClinux embedded operating system.An optimization method is proposed in order to reduce the total necessary cycle time and the algorithm redundancy consumed in real-time implementation. When the optimized G.723.1 algorithm encode a voice signal, the algorithm computation and the algorithm latency respectively are: under 6.3kbps is about 24MIPS and 12ms; under 5.3kbps is about 15MIPS and 8ms, these satisfies the technical specification and real-time requirements. Effect of the overall codec performance is about 10 times surperior to the traditional coding method and meet the real-time requirements of the BF 532 DSP platform with the uClinux embedded operating system , and all through the test with the ITU test vectors. Finally, the optimized G.723.1 algorithm applied to the embedded voice gateway, the experiments show that effects of voice calls is good.