| 语音识别中DTW特征匹配的改进算法研究 |
| |
| 引用本文： | 刘长明,任一峰.语音识别中DTW特征匹配的改进算法研究[J].中北大学学报,2006,27(1):37-40. |
| |
| 作者姓名： | 刘长明 任一峰 |
| |
| 作者单位： | 中北大学自动控制系,山西太原030051 |
| |
| 摘 要： | 采用基于动态规划方法的动态时间归正技术DTW(Dynamic Time Warping),可成功解决语音信号特征参数序列比较时时长不等的问题.在基于DTW的特征匹配用改进的动态时间归正方法将模板特征序列和语音特征序列进行匹配的基础上,比较两者之间的失真,得出识别判决的依据.实验表明,改进后的算法在孤立词语音识别中获得了良好性能.
|
| 关 键 词： | 动态规划(DP) 动态时间归正(DTW) 模板匹配 |
| 文章编号： | 1673-3193(2006)01-0037-04 |
| 修稿时间： | 2005年6月29日 |
A Research on Improving the Algorithm for DTW Feature-Matching in Speech Recognition |
| |
| LIU Chang-ming,REN Yi-feng.A Research on Improving the Algorithm for DTW Feature-Matching in Speech Recognition[J].Journal of North University of China,2006,27(1):37-40. |
| |
| Authors: | LIU Chang-ming REN Yi-feng |
| |
| Abstract: | The dynamic time warping(DTW) technology based on the dynamic programming can solve the problem that the characteristic parameters of speech signal are of different lengths in each other.Matching the pattern character and speech character with dynamic time warping enables the researchers to compare the distortions between them so as to obtain the criterion for identification.According to the experimental results,the algorithm after improvement performs better in single speech recognition. |
| |
| Keywords: | dynamic programming(DP) dynamic time warping(DTW) pattern matching |
| 本文献已被 CNKI 维普 万方数据 等数据库收录！ |
