维吾尔语陈述句韵律层级停顿模型研究

Research on Duration Modeling of Prosodic Levels in Uyghur Declarative Sentences

以提高合成语音自然度为研究背景,以建立陈述句中各韵律层级时长模型为研究目的,首先制定维吾尔语韵律层级划分规则,针对已筛选的40630个典型文本句子进行各韵律层级的人工标注,利用统计方法建立各个韵律层级的时长模型。实验结果表明,本文研究思路和方法切实可行,为提高语音合成自然度打下一个坚实的基础。

This paper takes the naturalness enhancement of TTS system as background,mainly studies the construction of duration models of prosodic layer in declarative sentences.At first,draws up the principals for boundary segmentation of prosodic layers in Uyghur sentences,then completes process of prosodic level tagging for 40630 representative sentences manually,at last builds up the duration model of each prosodic level.The research results show that the thoughts and methods for the particular problem in this paper are feasible and practicable,and it establishes a solid foundation for the naturalness enhancement of TTS system.