首页> 外文会议>Conference on Speech Technology and Human-Computer Dialogue >MODULE FOR GENERATING THE F0 CONTOUR USING AS INPUT A TEXT STRUCTURED BY PROSODIC INFORMATION
【24h】

MODULE FOR GENERATING THE F0 CONTOUR USING AS INPUT A TEXT STRUCTURED BY PROSODIC INFORMATION

机译:用于生成F0轮廓的模块,使用韵律信息结构的文本

获取原文

摘要

The paper presents a prosodic implementation for Romanian speech synthesis by applying a four-level intonational hierarchy model [7]. Using this hierarchy it follows good descriptions of intonational contour meaning. The input in this application consists in a text structured by an intonational unit tree, based on this hierarchy. The prosodic implementation involved the design of a rule-based submodule for F0 contour generation. The synthesised F0 contour and the phoneme sequence are directed to the input of a Klatt type synthesiser. In this paper, the intonational unit hierarchy is described by defining each intonational unit level. After a presentation of some F0 contour patterns corresponding to different types of phrasal events by which the phrasal units are delimited, a case study is analyzed in order to exemplify the results in the F0 frequency contour synthesis based on our intonation modelling. This application represents an improved implementation of intonation compared to that presented in [5], which is based on a two-level prosodic hierarchy only.
机译:本文通过应用四级语调层次模型来提出罗马尼亚语音合成的韵律实施[7]。使用此层次结构,它遵循语调轮廓的良好描述。本申请中的输入包括基于此层次结构由indonation单位树构造的文本。韵律实现涉及用于F0轮廓生成的基于规则的子模块的设计。合成的F0轮廓和音素序列涉及KLATT型合成器的输入。在本文中,通过定义每个钟表单元级来描述互动单元层次结构。在呈现与短语单元的不同类型的短语事件对应于不同类型的短语事件之后,分析了案例研究,以便基于我们的语调建模来举例说明F0频率轮廓合成的结果。该应用程序表示与[5]中的[5]中展示的相比改进了语调的实现,这仅基于两个级韵律层次结构。

著录项

相似文献

  • 外文文献
  • 中文文献
  • 专利
获取原文

客服邮箱:kefu@zhangqiaokeyan.com

京公网安备:11010802029741号 ICP备案号:京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有
  • 客服微信

  • 服务号