自回归预测多级矢量量化线谱频率编码技术

西安科技大学 通信与信息工程学院,陕西 西安 710054

线谱频率; 矢量量化; 码本设计; 自回归预测模型; 性能测试

Technology of multi-stage vector quantization with autoregressive prediction for linear spectrum frequency
CHEN Hui,ZHANG Bo-xia

(College of Communication and Information Engineering,Xi'an University of Science and Technology,Xi'an 710054,China)

linear spectrum frequency; vector quantization; design of codebook; autoregressive predictive model; performance testing

DOI: 10.13800/j.cnki.xakjdxxb.2017.0521

备注

在语音编码中线谱频率的量化编码多依赖于矢量量化技术。文中在分析经典的LBG多级矢量量化算法优缺点的基础上,结合m进制搜索代替全搜索以及瞬时联合调整各级码本的技术并引入自回归预测模型,实现了自回归预测多级联合矢量量化码本设计。并与窄带自适应多速率语音编码器AMR和MELP语音编码系统中线谱频率矢量量化进行了对比,效果良好。

In the speech coding,the quantization coding of the linear spectrum frequency(LSF)is mostly dependent on the vector quantization technique.Based on the analysis of the advantages and disadvantages of the classical LBG multistage vector quantization algorithm,this paper combines the m-search instead of the full search and the instantaneous joint adjustment of the codebook and introduces the auto-regressive prediction model to realize the design of auto-regressive prediction multistage joint vector quantitative codebook.Compared with the vector quantization of LSF in the narrow-band AMR and MELP speech coding systems,the effect is better.