论文部分内容阅读
语音的线性预测编码是获得低比特率数字语音的一种较好的方式.对预测滤波器模型参数采用矢量量化之后,将使这种方式更加完善,并在保证输出语音的质量下使比特率比一般线性预测进一步降低.本文在介绍了矢量量化的基本慨念之后,对采用矢量量化线性预测的几个主要问题如畸变测度的规定,产生码本的方法及搜索算法中的全搜索与二进位树搜索等作了扼要的阐述.本文最后介绍了一种较新的800bit/s矢量量化线性预测编码(LPC)声码器的主要结构.
Linear predictive coding of speech is a good way to obtain low bit rate digital speech.When vector quantization is used in predictive filter model parameters, this method will be more perfect, and the bit rate After the introduction of the basic concepts of vector quantization, this paper introduces several main problems of using vector quantization linear prediction, such as the distortion measure, the method of generating codebook and the full search in the search algorithm, Carry tree search, etc. In the end, this paper introduces a newer 800bit / s vector quantization linear predictive coding (LPC) vocoder of the main structure.