中文摘要：

本文首先给出了可训练语音合成系统架构，然后结合维吾尔语自身的特点，研究了其在维吾尔语音合成中的应用，给出了可训练语音合成系统训练部分所需数据准备．包括文本收集，录制、音素列表的确定、不带时间和带时间标注、问题集和上下文属性集的设计等。并利用通用工具HTS进行了测试，结果表明本文技术路线的可行性和所准备数据的有效性。

英文摘要：

In this thesis, the trainable text to speech synthesis system structure was described; then, an introduction on its appliance to the Uyghur speech synthesis was studied. Data preparation which is required for the training was described. The data preparing part includes the design of text collection, recording, mono labeling, context dependant labeling, design of question set and context attributes. The prepared data was sent to the HTS system and synthesized speech in Uyghur. The result indicates the availability of the technical way of this thesis and validity of the data prepared.