计算机科学 ›› 2014, Vol. 41 ›› Issue (1): 80-82.

• 2013 CCF人工智能会议 • 上一篇    下一篇

基于HMM的蒙古语语音合成技术研究

赵建东,高光来,飞龙   

  1. 内蒙古大学计算机学院 呼和浩特010021;内蒙古大学计算机学院 呼和浩特010021;内蒙古大学计算机学院 呼和浩特010021
  • 出版日期:2018-11-14 发布日期:2018-11-14
  • 基金资助:
    本文受国家自然科学基金项目(61263037),内蒙古自然科学基金重大项目(2011ZD11)资助

Research on HMM-based Mongolian Speech Synthesis

ZHAO Jian-dong,GAO Guang-lai and BAO Fei-long   

  • Online:2018-11-14 Published:2018-11-14

摘要: 基于隐马尔科夫模型的语音合成方法是当今语音合成的主流方法,它已被广泛应用于英语、汉语、日语等语音合成系统中。然而基于隐马尔科夫模型的蒙古语的语音合成技术研究还处于空白状态。首次将基于隐马尔科夫模型的语音合成方法用于蒙古语语音合成,并进行了语音合成实验。从最终合成系统的效果来看,合成的语音整体稳定流畅,可懂度高,而且节奏感比较强,主观平均得分为3.80。这为进一步研究基于隐马尔科夫模型的蒙古语语音合成技术奠定了基础。

关键词: 隐马尔科夫模型,蒙古语,标注,语音合成

Abstract: HMM-based speech synthesis method,as a mainstream method nowadays,has been widely applied to English,Chinese,Japanese,and so on.However,the research on HMM-based Mongolian speech synthesis is still in blank field.We applied the HMM-based speech synthesis method to Mongolian firstly,and did some experiments.From the evaluation results of the final Mongolian speech synthesis system,the synthesized Mongolian speech is stable,fluent,rhythmed and has high intelligibility.The mean opinion score of the synthesized Mongolian speech is 3.80.This laids the foundation for further research on the HMM-based speech synthesis technology.

Key words: HMM,Mongolian,Annotation,Speech synthesis

[1] 敖其尔,巩政.一种波形拼接的语音合成实验[C]∥第三届全国人机语音通讯学术会议.重庆,1994:408-412
[2] 萨其容贵.蒙古语语音合成技术的研究[D].呼和浩特:内蒙古大学,2005
[3] 田会利.基于词干词缀的有限条词的蒙古语语音合成系统的研究[D].呼和浩特:内蒙古大学,2007
[4] 孟和吉雅.基于动词词干词缀的蒙古语语音合成方法[J].内蒙古大学学报:自然科学版,2008,39(6):693-697
[5] 敖敏.基于韵律的蒙古语语音合成研究[D].呼和浩特:内蒙古大学,2012
[6] Zen Hei-ga,Takashi N,Junichi Y,et al.The HMM-basedSpeech Synthesis System (HTS) Version 2.0[C]∥6th ISCA Workshop on Speech Synthesis.Bonn,2007:294-299
[7] 井晓阳,罗飞,王亚棋.汉语语音合成技术综述[J].计算机科学,2012,9(11A),386-390
[8] 确精扎布,陈壮,何正安,等.GB 25914—2010传统蒙古文名义字符、变形显现字符和控制字符使用规则[S].北京,中国标准出版社,2010
[9] 清格尔泰.蒙古语语法[M].呼和浩特:内蒙古人民出版社,1991:65-66,6-77
[10] Tokuda K,Masuko T,Miyazaki N,et al.Hidden Markov models based on multi-space probability distribution for pitch pattern modeling[C]∥IEEE International Conference on Proceedings of the Acoustics,Speech,and Signal Processing.Arizona,1999:229-232
[11] masuko T,Tokuda K,Kobayashi T,et al.Speech synthesis from HMMs using dynamic features[C]∥IEEE International Conference on Proceedings of the Acoustics,Speech,and Signal Processing.Atlanta,1996:389-392
[12] Kawahara H,Masuda-Katsuse I,deCheveigne A.Restructuringspeech representations using pitch-adaptive time-frequency smoothing and an instantaneous-frequency-based F0extraction:possible role of a repetitive structure in sounds[J].Speech Communication,1999,27(3/4):187-207
[13] 吴义坚,王仁华.基于HMM的可训练中文语音合成[J].中文信息学报,2006,20(4):75-81
[14] Paul B,David W.Praat:doing phonetics by computer.http://www.fon.hum.uva.nl/praat/,2005
[15] CUED.Hidden Markov Model Toolkit (HTK).http://htk.eng.cam.ac.uk/,2009
[16] Satoshi I,Takao K.Speech Signal Processing Toolkit.http://sp-tk.sourceforge.net/,2012
[17] HTS working group.HMM-based Speech Synthesis System(HTS).2012.http://hts.sp.nitech.ac.jp/
[18] Wikipedia.Mean opinion score .http://en.wikipedia.org/wiki/Mean_opinion_score,2013

No related articles found!
Viewed
Full text


Abstract

Cited

  Shared   
  Discussed   
No Suggested Reading articles found!