Exploiting articulatory features for pitch accent detection

作     者：Junhong ZHAO Ji XU Wei-qiang ZHANG Hua YUAN Jia LIU Shanhong XIA 

作者机构：State Key Laboratory of Transducer TechnologyInstitute of ElectronicsChinese Academy of Sciences University of Chinese Academy of Sciences National Laboratory for Information Science and TechnologyDepartment of Electronic EngineeringTsinghua University 

出 版 物：《Journal of Zhejiang University-Science C(Computers and Electronics)》 (浙江大学学报C辑（计算机与电子（英文版）)

年 卷 期：2013年第14卷第11期

页      面：835-844页

核心收录：

学科分类：0711[理学-系统科学] 07[理学] 

基　　金：Project(Nos.61370034 61273268 and 61005019) supported by the National Natural Science Foundation of China 

主　　题：Articulatory features Pitch accent detection Prosody Computer-aided language learning(CALL) Multi-layer perceptron(MLP) 

摘      要：Articulatory features describe how articulators are involved in making *** often use a more exaggerated way to pronounce accented phonemes,so articulatory features can be helpful in pitch accent *** of using the actual articulatory features obtained by direct measurement of articulators,we use the posterior probabilities produced by multi-layer perceptrons(MLPs) as articulatory *** inputs of MLPs are frame-level acoustic features pre-processed using the split temporal context-2(STC-2) *** outputs are the posterior probabilities of a set of articulatory *** posterior probabilities are averaged piecewise within the range of syllables and eventually act as syllable-level articulatory *** work is the first to introduce articulatory features into pitch accent *** the articulatory features extracted in this way,together with other traditional acoustic features,can improve the accuracy of pitch accent detection by about 2%.