| 摘要: |
| 音高变化规律对合成高自然度的语音非常重要.为了提取音高即基频的变化规律，设计实现了数据挖掘系统SpeechDM，经过训练取得了较好的学习结果.该文介绍了国内进行的关于音高变化规则的研究情况以及数据挖掘技术，给出SpeechDM的系统结构，并介绍其主要模块，对实验结果进行了分析. |
| 关键词: 语音合成,音高变化,数据挖掘,神经网络,规则提取. |
| DOI： |
| 分类号: |
| 基金项目:本文研究得到国家自然科学基金和中国科学院百人工程项目基金资助. |
|
| Data Mining for Mandarin Tone Variation Patterns Learning |
|
ZHU Ting-shao,GAO Wen,Charles X. Ling,GAO Zheng-qi,LI Jin-tao
|
| Abstract: |
| The tone variation patterns is very important for speech synthesis. To extract more accurate patterns, a prototype system called SpeechDM has been implemented for learning. In this paper, the relative research and data mining are specified at first, the modules are introduced in detail, and some experimental results are also given. |
| Key words: Speech synthesis, tone variation, data mining, neural network, patterns extraction. |
