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。 In the research of Chinese speech recognition, the selection of recognition unit is very important. With the development of large vocabulary continuous speech recognition research, more and more Chinese speech recognition studies select the unit of syllable as the recognition unit. Combined with the phonetic characteristics of Chinese pronunciation, this paper proposes a strategy of overlapping phoneme segmentation based on syllables. The segmentation of syllables is realized by wavelet method. Experiments show that the method is accurate and reliable.