摘要
本文以一种定速率音段声码器采用的分段算法为基础,提出了一种以随机码本作为初始码本结合“联合分段量化法”的码本训练方法,改进了基音轨迹的表示方法,并验证了一种说话人自适应方法的效果。
Based on the segmentation algorithm of a fixed-rate segment vocoder, a new codebook training method is promoted in which the initial codebook is a random codebook and the joint segmentation and quantization algorithm is used. The quantization scheme of the pitch track is improved and a speaker adaptation method is also verified.
出处
《信号处理》
CSCD
2003年第5期448-452,共5页
Journal of Signal Processing
基金
中国科学院声学研究所知识创新工程前瞻项目基金