【作者】 库尔班·吾布力； 艾斯卡尔·艾木都拉；

【Author】 Kurban·Ubul Askar·Hamdulla Shchool information science and Engineering of Xinjiang University

【机构】 新疆大学信息科学与工程学院；

【摘要】 根据维吾尔语的特点,对Fujisaki量化模型进行了修改,使之用于分析维语语调, 以及它在通过规则合成F0曲线方面的应用。在已建立了的维吾尔语微型语音库的语音数据进行人工切分、标注得出手工词性标注提供的词类信息。通过分析数据库中的语句,研究了模型参数变化的原因。在此基础上构造了一套规则,能够捕捉语言和说话者相关的特征,并能为目标语句生成人工语调曲线。在一系列感知试验中测试了这一规则生成的维语语调模型的可接受性。最后,讨论这一方法规则的局限性。更多还原

【Abstract】 Fujisaki’s model is modified for analyzing the intonation of Uygur language, and its application on the regular synthesis of F0 contour based upon the characteristics of Uygur language. Phonetic data which getting from the speech database of Uygur language segmented and labeled manually, and obtained parts of speech information offered by handiwork labeling. The reasons of variations of the model’s parameter are studied by analyzing sentences in the database. And made a regulation, which able to capture related features of language and speaker, and generated man-made intonation curve for the target sentence. The acceptance of the Uygur intonation model generated by this regulation tested in a series of perceptive experiments. Finally, limitation of the rule is discussed.更多还原