论文部分内容阅读
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。
In order to improve the natural degree of synthesized speech, this study aims to establish the duration model of each prosodic hierarchy in the stated sentence. Firstly, the rule of Uyghur language hierarchy is established. According to the 40,030 typical text sentences that have been screened, Statistical methods to establish the duration model of each prosodic level. The experimental results show that the research ideas and methods in this paper are practical and can lay a solid foundation for improving the naturalness of speech synthesis.