| 一种基于分形理论的语音分割新方法 |
| |
| 引用本文： | 董远 胡光锐. 一种基于分形理论的语音分割新方法[J]. 上海交通大学学报, 1998, 32(4): 97-99 |
| |
| 作者姓名： | 董远 胡光锐 |
| |
| 作者单位： | 上海交通大学电子工程系 |
| |
| 摘 要： | 语音是由混沌的自然音素组成的，而分形可以很好地为成混沌状态的自然音素建模．语音波形具有分形特征，将分形用于改善语音识别技术越来越受到重视．语音的特性决定了每一个音素展现其固有模式，因此相邻音素之间的分维值不同．通常求取语音波形的分维值轨迹可把人的发音分割成句子、词、甚至音素．实验证明，该方法在语音分割中取得了很好的效果．
|
| 关 键 词： | 语音识别；语音分割；分形维数 |
New Way for Speech Segmentation Based on Fractal Theory |
| |
| Dong Yuan,Hu Guangrui,Sun Fang. New Way for Speech Segmentation Based on Fractal Theory[J]. Journal of Shanghai Jiaotong University, 1998, 32(4): 97-99 |
| |
| Authors: | Dong Yuan Hu Guangrui Sun Fang |
| |
| Affiliation: | Dong Yuan,Hu Guangrui,Sun Fang Department of Electronic Engineering,Shanghai Jiaotong University,China |
| |
| Abstract: | Speech utterances consist of chaotic natural phenomena, and fractal can model natural phenomena of chaotic state well. Speechwave appears in fractal feature. As the self pattern of each phoneme is determined by the character of utterances, the fractal dimension values between the phonemes are different. Speech utterance can be segmented into sentences, words, or even phonemes by the fractal dimension trajectory of speechwave. The experiment shows that the algorithm is effective in speech segmentation. |
| |
| Keywords: | |
| 本文献已被 CNKI 维普 等数据库收录！ |
|
