首页 | 本学科首页   官方微博 | 高级检索  
     

基于改进神经网络的三字词声调识别方法
引用本文:郑燕琳,许星宇,杨晓炯.基于改进神经网络的三字词声调识别方法[J].电声技术,2010,34(5):48-51.
作者姓名:郑燕琳  许星宇  杨晓炯
作者单位:公安部第三研究所,上海,200031
摘    要:汉语连续语音三字词声调由于受声道响应、音连、变调规律等的影响,情况较复杂,声调模式间交集大,模糊性强。提出了一种基于动态时间规整和改进神经网络的声调识别算法。实验表明,该算法可得到较好的识别效果,对探讨具有汉语特色的新的语音识别方法有重要意义。

关 键 词:连续语音三字词  声调识别  动态时间规整  改进神经网络算法

Tone Recognition of Three-syllabic Words Based on Improved Neural Network Algorithm
ZHENG Yan-lin,XU Xing-yu,YANG Xiao-jiong.Tone Recognition of Three-syllabic Words Based on Improved Neural Network Algorithm[J].Audio Engineering,2010,34(5):48-51.
Authors:ZHENG Yan-lin  XU Xing-yu  YANG Xiao-jiong
Affiliation:ZHENG Yan-lin,XU Xing-yu,YANG Xiao-jiong (The Third Research Institute of Ministry of Public Security,Shanghai 200031,China)
Abstract:The continuous Chinese tone of three-syllabic word is very complicated and its tone patterns lie in illegibility because of the effects of vocal response, syllable conjunction and tone sandhi. A method of tone recognition based on dynamic time alignment and improved neural network algorithm is presented. The simulation experiments show that the method can get a good recognition result. The method has an important meaning in exploring new speech recognition methods that possess Chinese characters.
Keywords:continuous three-syllabic words  tone recognition  dynamic time alignment  improved neural network algorithm  
本文献已被 CNKI 维普 万方数据 等数据库收录!
点击此处可从《电声技术》浏览原始摘要信息
点击此处可从《电声技术》下载全文
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