| 一种新的HMM训练方法 |
| |
| 引用本文： | 贺前华,陆以勤,韦岗.一种新的HMM训练方法[J].电子学报,2000,28(9):56-58. |
| |
| 作者姓名： | 贺前华 陆以勤 韦岗 |
| |
| 作者单位： | 华南理工大学电子与通信工程系,广州 510641 |
| |
| 基金项目： | 国家自然科学基金!(No.69881 0 0 1 )，广东省自然科学基金!(No .970 4 4 5) |
| |
| 摘 要： | 本文是对HMM最大距离训练方法的一种改进,该方法采用了更合理的模型距离定义,能更有效地利用训练数据集中的区别信息,使有限的训练数据得到更好的应用,达到提高语音识别系统性能的目的.导出了HMM模型参数的迭代公式.基于TIMIT数据库的非连续语音及连续语音实验结果表明,改进训练方法在降低错识率上较原来的方法有明显改善.
|
| 关 键 词： | 隐马尔可夫模型(HMM) 训练方法 判决信息 |
| 收稿时间： | 1999-04-21 |
A New Approach for HMM Training |
| |
| HE Qian-hua,LU Yi-qing,WEI Gang.A New Approach for HMM Training[J].Acta Electronica Sinica,2000,28(9):56-58. |
| |
| Authors: | HE Qian-hua LU Yi-qing WEI Gang |
| |
| Institution: | Dept.of Electronics Engineering,South China Univ.of Technology, Guangzhou 510641,China |
| |
| Abstract: | An improved maximum model distance approach was proposed to train HMM.By adopting a more realistic model distance definition,discriminative information contained in the training data could be used to improve the performance of recognizer.HMM parameter adjustment rules were induced.Both isolated word and continuous speech recognition experiments on TIMIT database showed that significant error reduction could be achieved by the improved approach. |
| |
| Keywords: | Hidden Markov Model (HMM) training approach discriminative information |
| 本文献已被 CNKI 维普 万方数据 等数据库收录！ |
| 点击此处可从《电子学报》浏览原始摘要信息 |
| 点击此处可从《电子学报》下载免费的PDF全文 |
