论文部分内容阅读
目前进行的基频变化规律提取主要是以人工总结为主 ,且大多只是给出定性的描述 ,这样很难得到完整全面的韵律规则及进行更深入的研究 .为了有效地进行韵律规则获取及进行更深入的研究 ,在已有实验的基础上利用聚类分析对零均值化后基频序列进行普通话句子中音节基频模式提取 ,取得了较好的实验结果 ,从而使得对基频变化规律的研究可以在定量的基础上进行 ,为基频规律的更深入研究提供了一条途径 .首先介绍有关背景知识以及进行基频模式发现的处理过程 ,然后介绍数据预处理技术以及所采用的聚类算法 ,最后给出实验结果
At present, the fundamental frequency variation rule extraction is mainly based on manual summarization, and most of them only give a qualitative description, so it is difficult to get a complete and comprehensive prosodic rule and carry out more in-depth research.In order to effectively carry out prosodic rule acquisition and conduct Based on the existing experiments, this paper uses the clustering analysis to extract the fundamental frequency of Mandarin syllables from zero-mean-valueized fundamental frequency sequence and obtains better experimental results, so that the change of fundamental frequency The research can be carried out on a quantitative basis and provides a way for deeper research on the fundamental frequency law.Firstly, the background knowledge and the processing of the fundamental frequency pattern discovery are introduced, and then the data preprocessing techniques and the clustering algorithm Finally, the experimental results are given