首页 | 本学科首页   官方微博 | 高级检索  
     检索      

基于二次多项式后验概率估计的汉语数码语音识别
引用本文:史媛媛,刘加.基于二次多项式后验概率估计的汉语数码语音识别[J].电路与系统学报,1999,4(4):36-41.
作者姓名:史媛媛  刘加
作者单位:清华大学电子工程系!北京,100084,清华大学电子工程系!北京,100084,清华大学电子工程系!北京,100084
基金项目:国家自然科学基金!69772020,国家863项目!863-512-9805-10
摘    要:本文提出了一种可在8位单片机上实现的非特定人汉语数码识别算法。该算法基于最大后验概率分类原理,利用二次多项式估计后验概率分布。其优点是识别算法的简单,不涉及动态规划过程。

关 键 词:汉语数码识别  后验概率  语音识别

Recognition of Mandarin Vocal numbers Based on the Estimation of Posteriori Probabilities using Quadric Polynomial
SHI Yuan-yuan, LIU Jia, LIU Run-sheng.Recognition of Mandarin Vocal numbers Based on the Estimation of Posteriori Probabilities using Quadric Polynomial[J].Journal of Circuits and Systems,1999,4(4):36-41.
Authors:SHI Yuan-yuan  LIU Jia  LIU Run-sheng
Abstract:A speaker independent Mandarin vocal number recognition algorithm that can be implemented in an 8-bitmicro-controller is proposed. The recognizer is based on the principle of classification by maximizing posteriori probability,and the a posteriori probabilities are estimated by a set of quadric polynomial functions. Wthout dealing with the dynamiccharacteristics of speech signal, the computing comPlexity of proposed recognition algorithm is very low. A recognizerconsisting of two cascaded functional levels and capable of selecting or rejecting is designed. Experiments prove that itsrecognition rate is high enough to satisfy the requirement of practical applications.
Keywords:Mandarin vocal number recognition  A posteriori probability  Polynomial estimation
本文献已被 CNKI 维普 等数据库收录!
设为首页 | 免责声明 | 关于勤云 | 加入收藏

Copyright©北京勤云科技发展有限公司  京ICP备09084417号