论文部分内容阅读
语音识别(Speech Recognition)应用的形式很多,目前主要分两大类:一是作为文稿数据的纯输入转换工具,如口述输入文稿或电子邮件等,其语音并不对机器实施控制,因此对语言“理解”的要求不高,重在正确地将输入转换为储存,器件所需词汇处理容量较大,如PC机中普遍应用的IBM语音输入软件,处理能力可达15万单词;二是对口述语言中所包含的要求或询问加以理解,作出正确响应,而不拘泥于所有词的正确转换,如手机语音拨号、TV音控选台、音控开关等,为减
There are many forms of application of Speech Recognition. At present, there are mainly two kinds of applications. One is pure input conversion tools such as spoken input manuscripts or e-mails that do not control the voice of the machine. Therefore, Understanding "is not high, focusing on the correct input is converted to storage, the device requires vocabulary processing capacity larger, such as PC commonly used in IBM voice input software, processing capacity up to 150,000 words; the second is the oral Language contained in the request or inquiry to understand and make the right response, and not rigidly adhere to the correct conversion of all words, such as mobile voice dialing, TV audio channel selection, audio switches, etc. to reduce