论文部分内容阅读
本文描述一个通用实时语言识别系统——RTSRS(01)。在以前工作的基础上,每条口呼命令的参数在时间域上规正,采用二值频谱,大大压缩了参考音的参数存贮量,同时应用新的求差距的办法,使得识别所需的时间大为缩短,以致字表为200时能实时识别。专人的识别结果为:口呼数字,99.7％;20句话(每句7字),99.7％;四字成语100个,99.5％;四字成语150个,99.3％;四字成语200个,98.8％;四字成语400个,99.7％。非正式的实验表明,对于不同音节数的字表,乃至口呼英语数字或BASIC语句名字等,都有高的正确识别率。
This article describes a universal real-time speech recognition system - RTSRS (01). On the basis of the previous work, the parameters of each command to be spoken are corrected in the time domain, and the binary spectrum is used to greatly reduce the parameter storage of the reference sound. At the same time, a new method of finding the difference is applied to make the identification Greatly reduced the time, resulting in 200 when the word table can be real-time identification. The identification result of the person is as follows: The number of the spoken words is 99.7%; 20 words (7 words per sentence), 99.7%; 100 words of the four words, 99.5%; 150 words of the four words, 99.3% 98.8%; four idioms 400, 99.7%. Informal experiments show that there is a high correct recognition rate for word lists of different syllables or even English numbers or BASIC statements.