基于高斯混合模型和残差预测的说话人转换系统  

A voice conversion system based on GMM and residual prediction

作　　者：吕声[1] 尹俊勋[1] 黄建成 

机构地区：[1]华南理工大学电子与信息学院,广东广州510641 [2]摩托罗拉中国研究中心,上海20041

出　　处：《电声技术》2004年第6期33-36,共4页Audio Engineering

摘　　要：说话人转换是将源说话人的语音特征转换成目标说话人的特征,使得听起来像是目标说话人的语音。提出的说话人转换系统分为2个部分,第一部分利用高斯混合模型进行谱包络的转换,训练采用时间对齐的源说话人和目标说话人的语音数据进行。第二部分基于一个分类器和残差码本对残差信号预测。该系统在现有的说话人转换系统的基础上做了一些改进,改进后不再需要说话人模仿别人的语调,并且在某些性能上超过了现有的系统。Voice conversion is the process of transforming the characteristics of speech uttered by a source speaker, such that a listener would believe that the speech was uttered by a target speaker. In this paper, the system is divided into two main parts. By using a Gaussian mixture model, which is trained on aligned speech from source and target speakers, the first part transforms the spectral envelope. The second part of the system predicts the spectral detail from the transformed LPC parameters, which is based on a classifier and residual codebooks. The system has some similarities with some existing systems, however, this system is not restricted to speech spoken in a monotone and with mimicked prosody. Also, on the basis of some performance metrics it outperforms existing systems.

关 键 词：说话人转换 高斯混合模型 残差预测 谱包络 

分 类 号：TN912.33[电子电信—通信与信息系统]