摘要
本文提出一种多特征汉语声调模糊识别方法,声调模式类别被描述为模糊集,从基音曲线上选取的三个特征物理量构成特征向量空间,在模糊空间中根据海明测度定义的模糊可信度被作为最终分类的准则。实验结果表明,这种方法的识别率高于99%。
In this paper, a multi-feature fuzzy recognition method for Chinese tone is proposed. The tone pattern classes are considered as fuzzy sets. Three features extracted from the pitch contours form the feature vector space. Every feature of each tone pattern class is also a fuzzy set. In the fuzzy space the Hamming distance is used to define the fuzzy belief, which makes the decision for final classification. The recognition rate is above 99% in our experiments.
出处
《模式识别与人工智能》
EI
CSCD
北大核心
1994年第1期60-65,共6页
Pattern Recognition and Artificial Intelligence
基金
国家自然科学基金
关键词
语音识别
元音识别
模糊识别
Speech Recognition, Chinese Tone, Fuzzy Belief.
