EI / SCOPUS / CSCD 收录

中文核心期刊

汉语语句重音对音高和音长的影响

The influence of Chinese sentence stress on pitch and duration

  • 摘要: 提高汉语合成语音的自然度的关键是要建立一个完善的汉语韵律模型.本文以连续的广播语言为研究对象,对汉语中语句重音对韵律特征参数的影响进行了初步探讨,分析了不同语句重音条件下音长和音高的变化及其相互关系,指出:(1)音高是语句重音的基本表达手段,随着语句重音级别的提高,音高分布曲线向高频方向推移。(2)在连续语流中词被"重读"和"轻读"的情况下,音长分布出现双峰,表示它们的音长有的受语句重音的影响,有的不受语句重音的影响。(3)在"正常"、"重读"和"轻读"王种情况下,音高和音长的相互关系分别是:不相关、负相关和正相关,证实了汉语语句重音音高和音长之间的互补关系。这些研究结果为汉语会成系统中韵律模型的建立提供了基础。在此基础上,本文又用神经网络对连续语音的语句重音进行了部分标注,开集中的分类结果正确率为63%,对语音数据中重音等级的自动标注方法作了探索。

     

    Abstract: The key for improving the naturalness of the syllthetic speech is to build a perfect prosodic model.Thepreliminary study of the influence of selitence stress on prosodic features in Mandarin was described.In continuousbroadcast speech,the relationship between duration and pitch under differellt sentence stress levels is analyzed.Theresults show that:(1) The pitch is the most important factor which carries the selltence stress,along with the increaseof the selltence stress level,the distribution colltour of pitch moves towards high frequency region.(2) In the case of "Stressed" and "Light" ,two peaks are found in the distribution colltour,which means that the duration of some wordsare affected by selltence stress,however others are not.(3) The relationship between pitch and duration in "Normal","Stressed "and" Light"are un-correlated,positive correlation and negative correlation respectively.The research laysa foundation for building a perfect prosodic model for Mandarin speech synthesis system.The results are also used toclassify selitence stress using a neural network,and 63% correction has been obtained in an open set test.A method oflabeling the sentence stress automatically was explored for Mandarin speech database.

     

/

返回文章
返回