王孟杰,孟子厚.基于语音参数的普通话韵母区别特征[J].声学技术,2011,(1):88~92 |
基于语音参数的普通话韵母区别特征 |
Research on distinctive features of mandarin finals based on speech parameters |
投稿时间:2010-01-30 修订日期:2010-04-07 |
DOI: |
中文关键词: 韵母 区别特征 语音参数 |
英文关键词: finals distinctive features speech parameters |
基金项目:江苏省高校自然科学研究计划项目,上海医疗器械高专科研启动项目 |
|
摘要点击次数: 1299 |
全文下载次数: 1462 |
中文摘要: |
现有区别特征理论缺少相应语音声学参数的支持,尝试从声学参数的角度来建立普通话韵母的区别特征。首先根据语音参数距离,使用层次聚类方法生成韵母决策树。在此基础上,按开尾韵和有尾韵的韵母分类方案,分析其聚类逻辑,提出了基于参数的普通话韵母区别特征。在非介音韵母中,区别特征层次首先根据主元音音位的高低区分;而在介音韵母中首先通过介音进行区分。使用支持向量机检测区别特征,韵母由区别特征节点组成的搜索路径所确定。通过在区别特征树上的搜索过程实现韵母的检测分类,其中非介音韵母的平均准确率为93.1%,介音韵母为92%。 |
英文摘要: |
There is little study about acoustic cues corresponding to the existing theory of distinctive features,this paper attempts to establish distinctive features of Mandarin finals based on speech parameters.Firstly,decision trees of Mandarin finals are constructed by using multi-layer clustering methods based on distances of speech parameters.According to finals classification scheme which include rhyme and open rhyme,and by analyzing the cluster logic of Mandarin finals’ decision tree,the distinctive features based on acoustic parameters are constructed.The distinctive features are arranged by the phoneme of main vowel firstly in finals without glides,and by the glides head in finals with glides.Distinctive features are checked by support vector machine and finals are identified according to a specific search path in distinctive feature trees.The accuracy of finals without glides is 93.1% and with glides is 92%. |
查看全文
查看/发表评论 下载PDF阅读器 |
关闭 |
|
|
|