摘要
文章讨论了高噪声背景下的汉语语音识别技术,从语音识别的一般方法出发,研究了基于半音节的半连续型隐马尔可夫模型(SCHMM)原理和算法,介绍了用DSP构建实现的方法,采用喉头送话器抑制了背景噪声,获得了较高的识别准确率。
A kind of Chinese speech recognition technology in loud ambient noise is discussed in this paper. Based on general speech recognition method, the principles and algorithms of semi-continuous Hidden Markov Model (SCHMM) are investigated. By introducing a throat MIC, the loud ambient noise is restrained, a higher recognition accuracy is achieved.
出处
《计算机与数字工程》
2009年第7期57-59,80,共4页
Computer & Digital Engineering
基金
"十五"国防预研基金资助课题