MELP语音编码器和AC-3音频解码器的研究及优化实现

【作者】 楼佳佳；

【导师】 刘云海；

【作者基本信息】 浙江大学 ， 信息与通信工程， 2007， 硕士

【摘要】 随着信息化社会的发展，对声音的压缩传递技术已经成为当今社会发展水平的一个重要标志，不论是用于人类通信的语音编解码技术，还是用于电视广播的音频编解码技术，都已成为当代社会生活不可或缺的重要组成部分，在多媒体、网络通信和保密通信等领域中发挥着各自的重要作用。混合激励线性预测(MELP)声码器由于其低码率、低时延、低复杂度等优点被广泛应用于商业、军事等方面的语音通信中，对其进行优化实现十分有现实意义。BF561是ADI公司新一代高性能的DSP，采用双核结构，内核时钟频率可达600MHz，且外围接口丰富，是多媒体应用的理想平台之一。本文将MELP编码器在BF561上进行实现及优化。AC-3音频标准也早在1995年就成为美国高清晰度电视标准组成部分之一，在电影、电视、广播等领域得到了广泛的应用，是大部分数字电视片源采用的声音压缩标准，在我国数字电视普及率越来越高的情况下，将此标准的解码算法进行实现和优化也有重大的现实意义。DM6446是TI公司新推出的集ARM与DSP于一体的双核DSP，非常适合多媒体应用，本文将AC-3解码器在DM6446的ARM上进行定点优化实现。论文首先对语音编码及音频编码的现状、发展以及编解码器的嵌入式应用现状进行了简要概述；其次介绍了语音编码基本原理，并依照MELP编码流程对各算法模块进行阐述；然后简要介绍了MELP编码系统及开发平台，包括BF561的结构特点、评估板的系统配置、软件平台Visual DSP等，并介绍了编码器输入输出接口的实现和系统运行时间的设置；接着阐述了在BF561平台上所使用的各种优化方法，主要介绍了MELP编码算法基于DSP的C代码优化、汇编代码优化及内存分配等，同时选取了几个例子加以说明；跟着详细介绍了AC-3音频解码算法、实现平台、系统框架和输入输出接口实现方法，将AC-3解码器浮点实现后进行定点化改造和存储空间优化，使其最终在DM6446的ARM上得以实现；最后对本文的工作进行了总结，并提出了今后的努力方向。更多还原

【Abstract】 With the development of informational society, the compression andtransmission technology of sound has already become important representation of thesocial development level. Both the speech coding/decoding technology, which is usedin human communication, and the audio coding/decoding technology, which is usedin TV and broadcast transmission, are important parts of the modem social life. Inadditional, these coding/decoding technologies play a very significant role in themultimedia, network communication and security communication.MELP(Mixed Excitation Linear Prediction) vocoder is widely applied incommercial and military speech communication because of its low-bit-rate, low-delayand low-complexity. Therefore, realizing the MELP speech coder is very useful andhas important meaning. BF561 is the new generation DSP produced by ADICorporation, which has two cores to deal with more instructions, and each core canwork at 600MHz. BF561 also has lots of peripheral interfaces. So, BF561 is an idealplatform for multimedia applications. In this thesis, we put the MELP speech coderonto this platform. Then we implement and optimize it on the BF561.The AC-3 audio standard became one part of the American HDTV standard in1995, and is widely applied in many fields, such as TV, film, broadcast etc. In China,digital TV is stepping in people’s normal life, in this situation, realizing the AC-3audio decoder also has significant practical meaning. DM6446 is newly proposeddual-core DSP by TI Corporation, which possesses both ARM and DSP properties.DM6446 is very suitable for multimedia applications. We apply AC-3 decodingalgorithm on the ARM of DM6446 in order to perform fixed-point description.The thesis at first surveys the development and the current progress of thespeech coding and current audio formats. Secondly, we introduce the principle ofspeech coder, and present the important model of MELP algorithm. Thirdly, wesimply introduce MELP codec system and its platform, including the structurecharacter of the BF561, audio sampling and sending on evaluating board, softwareplatform Visual DSP, and so on. Fourthly, we introduce methods to optimize programperformance and use them in MELP speech coder. Fifthly, we introduce the flow ofAC-3 decoding algorithm, realizing platform, system framework, I/O interface settingand its fixed-point description realization on the ARM of DM6446. Then we usefixed-point description to modify the program, and optimize the memory. Finally, wepresent the conclusion and tasks in the future.更多还原