摘要： 提出一种离散隐马尔科夫模型(hidden Markov model，HMM)和径向基函数(radial basis function,RBF)神经网络相结合应用于汉语数码语音识别(Mmandarin Ddigit Speech Recognition,MDSR)的方法。同时采用了一系列改进方法，使汉语数码语音的识别率达到了99.7%。

关键词: 汉语, 数码语音识别, 隐马尔科夫模型, 径向基函数

Abstract: This paper presents a new hybrid framework of hidden Markov model (HMM )and radial basis function (RBF) neural network for Chinese Mandarin digit speech recognition.Here, the HMM is employed to produce a best speech state sequence, which is warped to a fixed dimension vector and neural network is used as classifier.Results show that the new hybrid system works better than others.

中图分类号: 

• TN912.3