%0 Journal Article
%T 维吾尔语韵律建模<br>Prosody modeling for Uyghur TTS
%A 古力米热·依玛木
%A 姑丽加玛丽·麦麦提艾力
%A 玛依努尔·阿吾力提甫
%A 艾斯卡尔·艾木都拉
%J 清华大学学报（自然科学版）
%D 2017
%R 10.16511/j.cnki.qhdxxb.2017.21.026
%X 对维吾尔语的韵律结构进行了全面的研究，从维吾尔语语音合成（text to speech，TTS）语音库中提取了音节的时长、能量、基频均值、最大值、最小值和基频范围等韵律特征参数，分析了其在音节处于不同韵律层次时的变化规律。提取了语音数据中韵律边界前后的音节延长量、音高重置和无声段等声学特征参数，并对它们的分布规律进行了统计分析。实验结果表明：不同韵律层级之间时长延长量和音高差值随着边界层级的提高而增加；韵律词边界之间没有显著地停顿，韵律短语和语调短语层级边界之间的平均停顿时长分别是154.2和212.8 ms。<br>Abstract：The prosodic features of syllables such as duration, energy, mean pitch, maximum pitch, minimum pitch and pitch range were extracted from a Uyghur text to speech (TTS) database with analyses of their variations for different prosodic hierarchies. The pitch reset, pre-boundary lengthening, and silence duration of different prosodic boundaries were also analyzed. The results of acoustic experiments show that the pitch reset and pre-boundary lengthening are much greater as the prosodic boundary degree increases. No obvious pause can be perceived at the prosodic word (PW) boundary and the average silence duration at the prosodic phrase (PP) and intonation phrase (INP) boundaries are 154.2 and 212.8 ms.
%K 维吾尔语
%K 语音合成
%K 韵律结构
%K 声学特征分析
%K &lt
%K br&gt
%K Uyghur
%K text to speech (TTS)
%K prosody structure
%K acoustic analysis
%U http://jst.tsinghuajournals.com/CN/Y2017/V57/I12/1259