基于声学特征凸显的汉语疑问句检出

疑问句在语音交际中占有重要地位,检出语音交际中的疑问句对把握语义重心和分析话语结构具有重要作用。在分析疑问语气凸显的声学特征的基础上,提出了一种基于声学特征凸显的汉语疑问句检出方法。该方法用支持向量机(SVM)对小时间粒度的疑问语气凸显进行建模,通过基于加窗机制的后处理方法判决语音是否为疑问句。在面向电话客服应用环境录制模仿的和真实的汉语对话语音库上进行了算法测试。实验表明该方法对汉语疑问句具有较好的检出性能。

Acoustic features prominence based Chinese question detection

Question plays an important role in both human-human and human-computer speech communication.It is important to detect the questions for natural dialogues understanding and discourse structures analysis.The acoustic features of interrogative mood prominence in Chinese question are analyzed.A novel method is proposed to detect Chinese question based on acoustic fea-tures prominence.In particular,the support vector machine （SVM）is applied to model the interrogative mood prominence. Then,the windowed post-process is used to identify the question in the utterance.In the experiments,our method is applied to analyze both simulated and real-life Chinese corpora in the call center.The results demonstrate the effectiveness of our proposed model on Chinese question detection.