基于声韵母的汉语语音识别

Chinese Speech Recognition Based on Consonant and Vowel

随着大词汇量连续语音识别技术的发展,越来越多的研究人员选取声韵母作为识别单元。在基于声韵母的汉语连续语音识别中,声韵母基元的准确分割是非常重要的一步。结合汉语发音声学特性,提出了基于声母分割方法和基于段间距离方法相结合的策略。实验结果表明:该方法达到了准确分割的目的。

With the development of the technology of large vocabulary continuous speech recognition,more and more researcher select consonant and vowel as the recognition unit.Accurate segmentation of the consonant and vowel is a very important step in Chinese continuous speech recognition.Considering the acoustic properties of Chinese,this article proposes a new strategy that combines the method of dividing consonant and vowel segments with the method based on the distance between segments.The experimental results show that：the strategy can achieve the purpose of accurately division.