自适应回波抵消器和语音编解码器的设计实现

【作者】 杨乐；

【导师】 彭启琮；

【作者基本信息】 电子科技大学 ， 信号与信息处理， 2003， 硕士

【摘要】 本文的课题来自通信学院140教研室的预研课题——IP电话系统信号处理关键技术研究。主要工作是设计并实现电话线路自适应回波抵消器和基于多带激励模型的语音编解码器。对于IP电话系统而言，由于采用了分组交换技术，由2/4线转换器阻抗不匹配而产生的回波在系统中传输的时延较长，这将加重回波对语音通信的影响。所以，自适应回波抵消技术是IP电话系统不可或缺的技术之一。本文采用TMS320C5402 DSP实现了基于NLMS 算法和Geigle算法的回波抵消器。为了提高性能，本文引入了NLMS算法的扩展精度实现和Geigle 算法的高效实现。测试结果表明，回波抵消器的性能超出或达到了G.165建议标准的要求。语音压缩编码技术对提高IP电话系统的系统容量有重要的意义。基于多带激励模型的语音压缩编码算法以其较好的合成语音质量而受到广泛的重视。本文依据多带激励模型的定义，完成了编解码器算法设计，并用C语言实现。本文引入了一种新的基音周期计算方法，静音帧判决算法，清音帧判决算法，清浊音信息接收端重建等新算法，提高了合成语音的质量，降低了算法的总计算量。非正式编解码测试表明，本文的声码器能在2.4kbps的编码速率上合成出较好的语音。更多还原

【Abstract】 This work is supported by the pre-research project being conducted in 140 lab: " Researches on Key Signal Processing Technologies in IP Telephone Systems". And the main focus is the design, implementation and testing of an adaptive line echo canceller and a vocoder based on Multi-band Excitation (MBE) model.Because IP telephone systems employ the packet switching technology, the echo signals generated by the Hybrid due to the impedance mismatch usually experience longer transmitting delay, which significantly decrease the speech communication quality. Therefore, echo cancellers are indispensable devices for any IP telephone system. In this paper, an echo canceller based on NLMS algorithm and Geigle algorithm is implemented on TMS320C5402 Digital Signal Processor (DSP). In order to improve its performance, extended precision techniques and a high-efficiency implementation scheme of Geigle algorithm are employed. The testing results show that the performance of the echo canceller exceeds the requirements specified by ITU recommendation G.165.Speech compression technologies are of great importance in increasing the capacities of IP telephone systems. Vocoders based on the MBE model are popular due to their good synthesis speech quality. In this paper, a vocoder based on the MBE model is designed, its algorithms are detailed and it is implemented using C programming language. And a new pitch extraction algorithm, an active/inactive frame decision algorithm and a voiced/unvoiced frame decision algorithm are developed with the aims to improve the quality of the vocoder and reduce its overall computation load. The testing results show that the vocoder can synthesize high-quality speech when transmission rate is set to be 2.4kbps.更多还原