TONE ERRORS DETECTION BASED ON CONFIDENCE MEASURE*

摘要：

Mandarin is the tone language so that it is important to estimate the tone pronunciation of speaker in language learning. Tone variations in continuous speech can reduce the accuracy in detecting the tone errors in a computerassistant Mandarin language learning (CAMLL) system. In this paper the slopes of three uniformly divided F0 subsections are used as the features to construct the tone recognition models instead of traditional features F0 and F0, and the confidence measure is used for detecting the tone error. The results show the accuracy of the tones recognition and tone errors detection averagely improved 3％ and 10％.

关键词： computer aided language learning tone recognition tone errors detection confidence measure

作者: Ru Zhang Jiqing Han Jiu Lou

作者单位: School of Computer science and Technology,Harbin Institute of Technology,Harbin,Heilongjiang Province,China

会议类型: 国际会议

会议名称: 2008年拟人系统国际会议(2008 International Conference on Humanized Systems )(ICHS’08)

会议地点: 北京

会议语种:英文

在线出版日期: 2008-10-18（万方平台首次上网日期，不代表论文的发表时间）