| 欠通道混叠语音盲分离算法 |
| |
| 引用本文： | 郭水旺,吕凤英,陶玉福.欠通道混叠语音盲分离算法[J].河南科学,2010,28(4):440-443. |
| |
| 作者姓名： | 郭水旺 吕凤英 陶玉福 |
| |
| 作者单位： | 1. 黄淮学院,河南,驻马店,463000
2. 河南民权县地方税务局,河南,民权,476800 |
| |
| 基金项目： | 河南省科技发展计划项目 |
| |
| 摘 要： | 以超完备基的独立分量分析(Overcomplete ICA)为信道模型,采用自然梯度和最短路径的算法和提出的基于小波变换的Overcomplete ICA算法进行欠通道混叠语音盲分离,两者进行了比较.结果表明,该算法能够成功地分离混叠语音数据.
|
| 关 键 词： | 独立分量分析 自然梯度 最短路径 小波变换 |
Blind Source Separation of Overlapping Speech under Channel-Deficient |
| |
| Guo Shuiwang,Lv Fengying,Tao Yufu.Blind Source Separation of Overlapping Speech under Channel-Deficient[J].Henan Science,2010,28(4):440-443. |
| |
| Authors: | Guo Shuiwang Lv Fengying Tao Yufu |
| |
| Institution: | 1. Huanghuai University, Zhumadian 463000, Henan China; 2. Henan Minquan County Local Tax Bureau, Minquan 476800, Henan China) |
| |
| Abstract: | In practical separation of blind source speech, there is a phenomenon： the number of sensors usually less than the number of speech sources,which is refered to ＇separation of blind source speech under channeldeficient＇. To Overcomplete ICA as signal channel model, and propose a Overcomplete ICA algorithm based on wavelet translation which is used in＇separation of blind source speech under channel-deficient＇, and compare with previous natural gradient and shortest path algorithm. Experiments show this algorithm can separate overlapping speech date successfully. |
| |
| Keywords: | ICA natural gradient shortest path wavelet translation |
| 本文献已被 CNKI 维普 万方数据 等数据库收录！ |
|
