The Research Of AVS-P3 Decoding Algorithms And The Realization On DSP

In our country the industrial foundation on audio and video is good enough. However, because of the lack of the important techniques and standards, related corporations have long been enslaved to foreign corporations or organizations, which hold on the intellectual property. AVS audio is the third part of AVS which is put forward by out country independently. And it applies to the high-resolution digital broadcast, high-density laser digital storage medium, wireless broad-band multimedia communication and broad-band internet streaming media application.The purpose of this dissertation is to implement an efficient real-time AVS audio dual-channel decoder. First, the author introduces the core algorithms in AVS audio standards. AVS audio has the self-owned intellectual property rights. In order to be free of various audio patents, the audio standard introduces several new technologies, and a detail introduction of the high complexity modules is presented. Then by analyzing the tradeoff between performance and complexity of the core modules, some improvements on the reference decoder are introduced in the way of decoding speed and memory spending according to the analyzed result.The introduction of decoder realization on different DSP platform is divided into two parts. First, by analyzing the instruction feature of ZSP400 processor, the crucial modules are rewritten in assembly language. And the decoding performance is improved obviously, especially the computation-intensive codes. In the following chapter, the TIE of Xtensa HiFi2 Audio Engine is introduced to optimize the corresponding modules. The special instruction is quite effective in reducing the decoding complexity.Finally, according to the result of the consistence testing, the decoder could decode all the testing streams except for the format of AASF and 5.1 channels streams, and the subjective quality of the decoded sample is good. With most of general testing stream, the difference of the decoded sample value of real-time decoder and the reference decoder is no more than 20. The performance of the decoder is 50M Hz on ZSP400 platform, and 110M Hz on Xtensa HiFi2 Audio Engine platform. The research of this dissertation is useful for the spread and application of AVS audio.