改进的混合激励线性预测编解码器算法

发布时间：2018-03-05 11:21

本文选题：语音　切入点：声码器　出处：《青海师范大学》2015年硕士论文　论文类型：学位论文

【摘要】：随着科技的发展,如何在众多通信系统中有效的传递语音信息仍然是人们研究的重点。最早的语音通信采用模拟语音通信实现了语音的远距离传输,但随后出现的数字语音通信,在可靠性、快速交换、抗干扰等方面以无法比拟的优势取代了模拟语音通信,它的缺点是过多的占用了频带。但通过语音编码技术的深入研究,可以有效的降低信号传输时占用的频带。近年来,语音通信带宽有限的问题似乎已经得到解决,但是随着数字通信业务的发展和实际通信中的需要,继续降低语音编码速率,提高信道利用率成为了人们的迫切需要。因此较高质量的低速率语音编码算法成为了语音编码技术中必不可少的研究方向。目前,低速率语音编码技术在算法和结构这两个方向上的发展变得越来越复杂,混合激励线性预测(MELP)算法作为其中的优秀算法之一,是目前低速率语音编码中最具潜力的算法。该算法在基本的LPC模型基础上,用混合激励取代过于简单的二元激励,利用多带思想,更好的模拟出自然语音特性,合成出较高质量的语音。本文经过对MELP算法编解码原理的分析,针对MELP声码器基音周期提取的复杂性和子带声音强度矢量量化粗造性,提出了两项大的改进。一是将基音周期的提取方法改进成一种较为简化的算法；二是对子带声音强度矢量量化方法的设计和实现,得到了一种改进的MELP声码器算法(A-MELP)。最后对改进的MELP算法进行了大量、充分的仿真实验。
[Abstract]:With the development of science and technology, how to transmit voice information effectively in many communication systems is still the focus of research. But the subsequent digital voice communication replaces analog speech communication with incomparable advantages in reliability, fast switching, anti-jamming and so on. Its shortcoming is that it occupies too much frequency band. But through the in-depth study of speech coding technology, In recent years, the problem of limited bandwidth of voice communication seems to have been solved, but with the development of digital communication service and the need of actual communication, the rate of speech coding continues to decrease. Improving channel utilization has become an urgent need. Therefore, high quality low-rate speech coding algorithm has become an indispensable research direction in speech coding technology. The development of low rate speech coding technology in the direction of algorithm and structure is becoming more and more complicated. The hybrid excited linear predictive algorithm is one of the excellent algorithms. It is the most potential algorithm in low rate speech coding at present. Based on the basic LPC model, the algorithm uses mixed excitation to replace the simple binary excitation, and makes use of the multi-band idea to better simulate the natural speech characteristics. A high quality speech is synthesized. Based on the analysis of the encoding and decoding principle of MELP algorithm, the complexity of pitch period extraction of MELP vocoder and the coarseness of subband sound intensity vector quantization are analyzed. Two major improvements are proposed. One is to improve the pitch period extraction method into a simplified algorithm, the other is to design and implement the subband sound intensity vector quantization method. An improved MELP vocoder algorithm is obtained. At last, a large number of simulation experiments are carried out on the improved MELP algorithm.
【学位授予单位】：青海师范大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TN76

【相似文献】