基于模糊粗神经网络的汉语声韵母切分  

I/F Segmentation for Mandarin Speech Based on Fuzzy-rough Neural Network

作　　者：苏敏[1] 李战明[1] 

机构地区：[1]兰州理工大学电气与信息工程学院,甘肃兰州730050

出　　处：《电声技术》2008年第9期49-51,共3页Audio Engineering

摘　　要：针对汉语连续语音,提出一种声韵母切分方法。根据扩展的声韵母为识别基元,采用汉语音节的重叠音素分割策略,利用模糊粗神经网络进行声韵母自动切分。实验室实验证明了该方法进行音节分割的有效性和合理性。A new algorithm for mandarin continuous speech I/F（Initial and Final） segmentation is presented. Based on the XIF（eXtend Initial/Final） set chosen as the basic speech recognition unit, overlapped segmenting scheme is used, and fuzzy-rough neural network is used to perform the phoneme segmentation. The validity and accuracy of the method are proved by experiments.

关 键 词：语音识别 声韵母 重叠音素分割 模糊粗神经网络 

分 类 号：TN912[电子电信—通信与信息系统]