基于说话人自适应训练的汉藏双语语音合成  

Realizing Mandarin-Tibetan bilingual speech synthesis by speaker adaptive training

作　　者：王海燕[1] 杨鸿武[1] 甘振业[1] 裴东[1] 

机构地区：[1]西北师范大学物理与电子工程学院,兰州730070

出　　处：《清华大学学报（自然科学版）》2013年第6期776-780,共5页Journal of Tsinghua University(Science and Technology)

基　　金：国家自然科学基金项目(61263036;61262055);甘肃省杰出青年基金项目(1210RJDA007)

摘　　要：根据藏语和汉语在发音上的相似性,提出了一种基于隐Markov模型(hidden Markov model,HMM)的汉藏双语语音合成方法。以声韵母为合成基元,采用多个普通话说话人和1个藏语说话人的语料库,利用说话人自适应训练,获得一个汉藏双语混合语言的平均音模型。通过说话人自适应变换,从混合语言的平均音模型获得普通话或藏语的说话人相关模型,从而合成出普通话或藏语语音。实验结果表明,在藏语训练语句较少的情况下,该方法合成的藏语语音明显优于仅采用说话人相关模型合成的藏语语音。This paper presents a method to realize hidden Markov model（HMM）-based Mandarin-Tibetan bilingual speech synthesis using the similarities between Mandarin and Tibetan pronunciation.The initial and the final are used as the synthesis units with training using a set of average mixed-lingual models from a large Mandarin multi-speaker-based corpus and a small Tibetan one-speaker-based corpus using speaker adaptive training（SAT）.Then,the speaker adaptation transformation is applied to the speaker dependent（SD） training data to obtain a set of speaker dependent Mandarin or Tibetan models from the average mixed-lingual models.The Mandarin speech or Tibetan speech is then synthesized from the speaker dependent Mandarin or Tibetan models.Tests show that this method outperforms the method using only Tibetan SD models when only a small number of Tibetan training utterances are available.

关 键 词：语音合成 隐Markov模型(HMM) 说话人自适应训练 多语种语音合成 藏语语音合成 汉藏双语语音合成 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]