【摘要】 普通话是有调语言,基频是TTS系统中选择单元时一个非常重要的参数。为了能根据基频这个声学参数来选择语音单元,就必须建立文本上下文环境信息与基频曲线之间映射关系,即基频模型。本文将通过决策树的方法来提取这个模型,并将这上模型应用到普通话的文—语转换系统中。更多还原

【Abstract】 Mandarin is a language with five tones.Pitch is a very important parameter for unit-selection in Mandarin TTS.We should find the mapping between text context and the pitch contour to improve unit-selection.This mapping is called pitch-mod- el.In this paper,a method by using decision tree will be performed to generate this model.This model will be applied in man- darin text-to-speech system.更多还原