| 连续汉语音流中声韵母切分研究 |
| |
| 引用本文： | 刘映杰,马义德,刘悦,袁敏,段磊.连续汉语音流中声韵母切分研究[J].甘肃科学学报,2005,17(3):90-93. |
| |
| 作者姓名： | 刘映杰 马义德 刘悦 袁敏 段磊 |
| |
| 作者单位： | 兰州大学,信息科学与工程学院,甘肃,兰州,730000 |
| |
| 基金项目： | 甘肃省自然科学基金暨中青年科技基金（YS021-A22-009）;兰州大学985特色项目基金 |
| |
| 摘 要： | 根据汉语语音的特点,提出了一种连续汉语语音中声韵母自动切分的新算法.根据音素单元结构确定HMM模型中的状态数,利用HMM模型进行音素切分,再对切分结果进行合并,得到最终的声韵母子词单元.通过实验验证了子词单元产生的稳定性和准确性,实验结果表明,算法在这两方面均达到了较好的性能,且与话者无关.
|
| 关 键 词： | 语音识别 子词 切分 HMM模型 |
| 文章编号： | 1004-0366（2005）03-0090-04 |
| 收稿时间： | 2004-10-25 |
| 修稿时间： | 2004年10月25 |
Research on I/F Segmentation in Continuous Mandarin Speech |
| |
| LIU Ying-jie,MA Yi-de,LIU Yue,YUAN Min,DUAN Lei.Research on I/F Segmentation in Continuous Mandarin Speech[J].Journal of Gansu Sciences,2005,17(3):90-93. |
| |
| Authors: | LIU Ying-jie MA Yi-de LIU Yue YUAN Min DUAN Lei |
| |
| Abstract: | A new algorithm of segmenting automatically initial/final subword units is presented based on the characteristics of mandarin speech. The number of states in HMM is confirmed in terms of phoneme unit structures. Phoneme segmentation is performed using HMM and the segmentation results are combined. The segmentation of initial/final subword units is obtained. The robustness and accuracy of generating subword units are proved by experiments. The results show that the algorithm has a better performance on the two aspects, and is independent of speakers. |
| |
| Keywords: | speech recognition subword segmentation HMM model |
| 本文献已被 CNKI 维普 万方数据 等数据库收录！ |
|
