连接数字语音识别  

Connected Digit Speech Recognition

作　　者：张春涛[1] 吴善培[1] 

机构地区：[1]北京邮电大学电信工程学院

出　　处：《北京邮电大学学报》1997年第4期49-53,共5页Journal of Beijing University of Posts and Telecommunications

摘　　要：采用分段Ｋ均值法进行模型的训练，采用分级构造的方法进行连接数字的识别．在训练和识别中，又引入了状态持续时间模型．并且对易混淆的字采用声韵分割的方法训练模型，得到了较好的识别率．A segmental K means training is used to train models and a level building method is used to recognize connected digit. State duration model is also used during the training and recognition procedure. Partition of the Mandarin syllable into the initial part and the final one is applied to train the models of confusing digits,good recognition rate is reached.

关 键 词：语声信号处理 语音识别 声韵分割 

分 类 号：TN912.34[电子电信—通信与信息系统]