摘要
汉语是一种带声调的语言,声调信息在汉语识别中具有非常重要的意义.传统的声调识别一般只研究朗读式语音中相对标准的声调,很少对声调调型比较复杂的自然口语进行专门的处理.针对汉语自然口语的特点,在声调建模单元的选择时提出了真实上下文的模型.同时,为了对声调模式进行精细建模,采用了一种层次聚类的方法来获得更多的声调模式.实验结果证明了方法的有效性.
Chinese is a tonal language, and the tone information is very important for the recognition of Chinese speech. Conventional methods generally focus on the normal tone in reading speech, rather than on the complicated spontaneous speech specifically. The real-context model is proposed as a new concept to be used in the tone unit selection. Then a method based on hierarchical clustering is performed to generate a more refined tone model. Experimental results have proved the effectiveness of the methods.
出处
《物理学报》
SCIE
EI
CAS
CSCD
北大核心
2007年第12期7064-7069,共6页
Acta Physica Sinica
基金
国家重点基础研究发展规划(批准号:2004CB318106)
国家自然科学基金(批准号:10574140
60535030)资助的课题.~~
关键词
声调识别
自然口语
真实上下文模型
聚类
tone recognition, spontaneous speech, real-context model, clustering
