【摘要】 焦点是语言表达的重要方式,焦点重音是重要的韵律特征,实现中性语音到焦点语音的转换可以提高语音的表现力。该文提出了声学特征局部凸显度的表示方法,分析了由中性语音到焦点语音,焦点单词所属音节声学特征变化与中性语音相应音节声学特征局部凸显度的相关性,提出了一种基于决策树的英语焦点语音的转换模型。该模型采用决策树对训练语料进行聚类,所用上下文包括音节与焦点单词的相对位置以及音节在韵律结构(如韵律短语、韵律词等)中的位置。在此基础上,提出了一种基于局部凸显度的中性语音到焦点语音声学特征变化的预测算法。采用该算法后,客观实验中声学特征变化平均绝对值误差降低到0.08,主观实验表明本文提出的模型的转换语音具有更好的焦点表达效果和自然度。更多还原

【Abstract】 Emphasis is an important feature of prosody. Emphatic speech conversion techniques can improve the expressiveness of converted speech. This paper defines the local prominences( LPs) of acoustic features and then analyzes the correlations between the feature changes of the syllables of the emphasized words from neutral to emphatic speech and the LPs of the corresponding syllables of the neutral speech. An English emphatic speech conversion model is then developed based on a decision tree. This model uses the decision tree to cluster the training data. After this data clustering,an algorithm is given to predict feature changes from neutral to emphatic speech based on the LPs. Tests demonstrate that the mean average error of the model is decreased to 0. 08. The naturalness and the emphasis intensity of the converted speeches are also improved.更多还原