Implementation And Optimization Of The AMR-WB Algorithm Based On DM642

With the development of 3G (the third generation mobile communication), to satisfy the requirements for voice service of high quality and make up for the defects of current narrowband's voice coding, 3GPP/ETSI put forward a AMR-WB in 2001, which has 9 CODEC patterns in different rates with low average rate of coding and good adaptability. The phonosensitive quality is improved with its frequency band range expanding to 50-7000Hz and the frequency of sampled signal being 16kHz. In the meantime, AMR-WB is the first audio coding standard which applies to both wired and wireless networks.Adopting the ACELP model, AMR-WB is of high computational complexity. The operation of the codes of standard protocol provided by 3GPP is quite time-consuming, therefore, the corresponding optimization processing need to be done according to specific platform while in practice. In this paper, the principe of AMR-WB algorithm is studied and each module is analysed;A new hardware platform ,realize the AMR-WB algorithm,is introduced through studying the hardware resources of the DM642. the ways in terms of optimization and realization of AMR-WB algorithm is also studied. The optimization processing includes: the macro definition of inline function of the basic operators; the realization and optimization for the most time-consuming parts of key modules in linear assembly way. Data alignment could be reached by opening up stack space to make full use of packet bundling and software pipelining of DSP. This paper researches the specific application of each method that optimizes C language and assembly in AMR-WB algorithm based on compiler. The paper also studies the interface standard of mixed programming based on C language and linear assembly; analyzes the process of debugging and test of AMR-WB's CODEC software, and finds the ways that enable the algorithm to implement in real time on the DM642's hardware platform.The test and analysis of the results before and after the optimization of AMR-WB algorithm show: after the optimization, the algorithm has been greatly improved in terms of computing time and storage area, and the voice quality in output testing is consistent with the one in standard testing.