Implementation Of AMR Speech Codec Based On Tilera Processor

Adaptive multi-rate(AMR) is a speech coding standard which applies to the third generation mobile communication system WCDMA. AMR speech coding established by 3GPP is a variable speech coding technology which is implemented with the algorithm of algebraic code excited linear prediction and supports up to 8 coding rates for speech signal sources. AMR speech codec can choose the best channel mode and coding mode adaptively according to the conditions of the wireless transmission channel. It is widely used in the mobile communication now.AMR speech coding has large computational complexity. So it is necessary to implement AMR on high-performance hardware devices and optimize it in order to improve real-time voice communication. Therefore, improving the key technologies of AMR algorithm and researching to the hardware realization platform have important practical value and a good prospect of application.Now Multi-core processor chips as well as high-performance application systems develop very rapidly. Tilera processors are high-performance processors which are developed by Tilera company. They are good at high speed, parallel communications and data processing. What'more, tilera processor has a unique capacity for concurrent processing in digital signal processing, stream processing, and network security. Of course Tilera processor can also improve real-time voice communication.Firstly the thesis studies the principle and critical techniques of AMR algorithm,and analyzes the architecture and function of Tilera multi-core processor as well as the utilization of multi-core development environment MDE. Secondly, AMR speech codec is ported to TilePro64 multi-core platform according to the structural characteristics of TilePro64 processor.We optimize it simply on TilePro64 platform. Finally, the performance of AMR speech codec after optimization is tested with ITU standard voice sequences. The experiment results show that the efficiency of AMR speech codec has greatly improved.