摘要
语音识别中，动态时间规整和隐马尔可夫划最有效的两种识别算法，并且ＤＴＷ和ＨＭＭ在本质上是一致的。根据ＤＴＷ和ＨＭＭ的本质联系和各自所对应的声学模型，在前期工作中建立了一种广义声学模型，并指出ＤＴＷ和ＨＭＭ只是ＧＭ的特例，且ＤＴＷ和ＨＭＭ都可以转化为ＧＭ。
In previous work,a General model(GM) had been established to describe the model of speech.Based on this,this paper proposes an algorithm of GM by introducing Fisher algorithm,at the same time,the convergence of GM has also been analyzed,and finally,the conclusion is drawn that the algorithm of GM is convergent by probability.
出处
《计算机工程与应用》
CSCD
北大核心
2000年第2期60-62,共3页
Computer Engineering and Applications
