| 改进型重复Wiener滤波/PUM模型--实现抗噪连续语音识别 |
| |
| 引用本文： | 林劼,Ji Ming,刘玓. 改进型重复Wiener滤波/PUM模型--实现抗噪连续语音识别[J]. 计算机工程与应用, 2006, 42(11): 41-44,79 |
| |
| 作者姓名： | 林劼 Ji Ming 刘玓 |
| |
| 作者单位： | 电子科技大学计算机科学与工程学院,成都,610054;英国贝尔法斯特女王大学计算机科学系 |
| |
| 摘 要： | 众所周知,抗噪问题是现在语音识别研究中的重点。文章描述了一种新的抗噪语音识别方法,即通过改进型重复Wiener滤波结合后验概率联合模型PUM(PosteriorUnionModel)[3]实现在噪声环境下连续字语音识别的方法。这种方法先采用改进型重复Wiener滤波器对语音信号进行语音增强预处理,消除已知噪声,为PUM模型提供只有局部频带被噪声污染的语音信号,再利用PUM模型进行抗噪语音识别。试验表明在各种不同的噪声环境下新方法有更高的平均识别率。
|
| 关 键 词： | 重复 Wiener 滤波 PUM模型 抗噪语音识别 |
| 文章编号： | 1002-8331-（2006）11-0041-04 |
| 收稿时间： | 2005-07-01 |
| 修稿时间： | 2005-07-01 |
Improved Iterative Wiener Filters and Posterior Union Model Robust Speech Recognition |
| |
| Lin Jie,Ji Ming,Liu Di. Improved Iterative Wiener Filters and Posterior Union Model Robust Speech Recognition[J]. Computer Engineering and Applications, 2006, 42(11): 41-44,79 |
| |
| Authors: | Lin Jie Ji Ming Liu Di |
| |
| Abstract: | In this paper we present a new approach towards robust automatic speech recognition in adverse conditions. This new method is based on the combination of a speech enhancement using Improved Iterative Wiener Filters,and Posterior Union Model(PUM)[3] to dynamically modify the probability computations performed in GMM recognizers. Previous work has demonstrated that the PUM succeeds in robust speech recognition with some interruptions and band-limited noises. In this paper,we show that this new method more improves ratio of speech recognition in wide-band noisy conditions. |
| |
| Keywords: | Wiener |
| 本文献已被 CNKI 维普 万方数据 等数据库收录！ |
