QQ客服 官方微博 反馈留言

题名	低速率多带激励语音编码算法研究
作者	杨明
学位类别	博士
答辩日期	2000
授予单位	中国科学院声学研究所
授予地点	中国科学院声学研究所
关键词	语音编码 多带激励 基音提取 共振峰 线性预测
中文摘要	在数字通信中，语音信号直接数字化所需的数码率太高，为了提高传输和存储的效率，充分利用信道容量，必须对数字语音信号进行压缩编码。通过降低编码速率，可以使同样的信道容量能够传输更多路的语音信号。在传输比特率限制十分严格的场合，低速率语音编码具有特别重要的意义。本文对多带激励（MBE）算法进行了深入研究，并以MBE-LPC编码方案为基础，用C语言模拟实现了一个速率为2.4kb/s的语音编码/解码系统。合成语音质量比较令人满意，可以满足一般的通信需要。基音频率是MBE模型中最重要的模型参数，对系统整体质量起着决定性的影响。本文采用了一种有效的基音提取和平滑算法，准确度较高，并注重消除了倍频干扰，具有良好的基音连续性，提高了合成语音的自然度。针对现有模型在声道响应上的缺陷，本文提出了一种新的语音编码方案，主要在合成方法上进行了调整。基本思想是利用多带激励的概念合成残差信号，再采用LPC综合滤波得到最终的合成语音。以此方案建立的语音编码/解码系统传输速率降到了1.6kb/s。经重建语音信号比较及主观试听表明，该系统性能与美国联邦标准推荐的2.4kb/s混合激励线性预测（MELP）算法接近或相当。
英文摘要	In digital communication, the bit rate is too high when speech signals are directly quantized. In order to improve the efficiency of transmission and storage, digital speech signals must be compressed and coded so that the channel capacity can be more efficiently used. By reducing coding rate, more speech signals can be transferred in the same channel. At the situation where the transmission rate is limited very strict, Low bit Rate Speech Coding has especially important significance. In this thesis, Multi-Band Excitation (MBE) algorithm is deeply studied. A speech coding/decoding system at 2.4kb/s based on MBE-LPC coding method is simulated by language C. The quality of the synthesized speech is nearly satisfied and can meet most need in communications. Fundamental period is the most important parameter in MBE model, which definitively influence the quality of the whole system. An efficient pitch extracting and tracking algorithm is implemented in the thesis. It can obtain relatively accurate pitch estimate, and pay more attention to reduce multiple frequency interfering. By this method, the naturalness of the synthesized speech is improved. In order to overcome the deficiency in the vocal track response of the present model, the thesis proposed a novel speech coding method, whose adjustment is mostly in the synthesis technique. The basic idea of this method is to get synthesized residue using the concept of Multi-Band Excitation, and then pass a LPC synthesis filter to obtain the ultimate synthesized speech. The transmission bit rate of the coding/decoding system based on this method is reduced to 1.6kb/s. By comparing reconstructed speech signals and subjective evaluation, it is concluded that the performance of this system approaches or corresponds to that of 2.4kb/s Mixed Excitation Linear Prediction (MELP) algorithm which is in Federal Telecommunication Recommendation.
语种	中文
公开日期	2011-05-07
页码	66
内容类型	学位论文
源URL	[http://159.226.59.140/handle/311008/690]
专题	声学研究所_声学所博硕士学位论文_1981-2009博硕士学位论文
推荐引用方式 GB/T 7714	杨明. 低速率多带激励语音编码算法研究[D]. 中国科学院声学研究所. 中国科学院声学研究所. 2000.

个性服务

查看访问统计

相关权益政策

暂无数据

收藏/分享

除非特别说明，本系统中所有内容都受版权保护，并保留所有权利。