论文部分内容阅读
文章介绍了用HMM嵌入训练方法来建立连续语音的声学模型，并对基于音素的HMM和基于音节的HMM进行了比较，并以此为基础建立了一个实用的银行电话语音服务系统。
In this paper, an acoustic model of continuous speech based on HMM embedding training is introduced. The comparison between HMM based on phonemes and HMM based on syllables is also presented. Based on this, a practical banking telephone voice service system is established.