摘要
汉语语音识别研究中,识别单元的选取是很重要的。随着大词汇量连续语音识别研究的发展,越来越多汉语语音识别研究中选取次音节单位作为识别单元。结合汉语发音声学特性,提出了音节的重叠音素分割策略,并利用小波方法实现了音节的分割,实验证明该方法分割准确可靠。
It's very important to choose speech recognition units in Mandarin speech recognition researches.As the development of large vocabulary continuous speech recognition,sub--syllable is used as recognition units in many Mandarin speech recognition researches.In this paper,based on Mandarin acoustic characteristics,an overlapped segmenting scheme is presented and wavelets are used to segment phoneme.Experiments show that the method is available.
出处
《火力与指挥控制》
CSCD
北大核心
2004年第6期94-96,99,共4页
Fire Control & Command Control
关键词
音素
语音识别
小波变换
突变检测
phoneme,speech recognition,wavelets transform,jump-detection
