论文部分内容阅读
本文提出了基于遗传算法(GA)的矢量量化话者模型,简称GAVQ话者模型。给出了建立该模型的算法即GAVQ算法。矢量量化技术可以有效地提取和表征话者的个性信息,因而VQ码本可以有效地用作话者模型,传统的VQ码本设计方法是LBG算法,但该算法是一种局部优化算法。GAVQ算法将遗传算法的全局优化特性和VQ建模技术巧妙地结合起来,采用科学的编码方案,动态的定标技术,高效的交叉策略,是一种全局优化的vQ码本设计算法。并进行了实验检验。
This paper presents a vector quantization speaker model based on genetic algorithm (GA), referred to as GAVQ speaker model. GAVQ algorithm is given to build the model. VQ codebook can effectively be used as the speaker model. The traditional VQ codebook design method is LBG algorithm, but the algorithm is a local optimization algorithm. Vector quantization can effectively extract and characterize the speaker’s personality information. GAVQ algorithm combines the global optimization of genetic algorithm and VQ modeling technology skillfully. It adopts a scientific coding scheme, dynamic scaling technology and efficient cross strategy. It is a globally optimized vQ codebook design algorithm. And the experimental test.