| 基于噪声短时谱动态估计的语音增强谱减算法 |
| |
| 引用本文： | 李银国,薛雯,徐洋.基于噪声短时谱动态估计的语音增强谱减算法[J].重庆邮电大学学报(自然科学版),2010,22(2):127-130. |
| |
| 作者姓名： | 李银国 薛雯 徐洋 |
| |
| 作者单位： | 重庆邮电大学自动化学院,重庆,400065;重庆邮电大学自动化学院,重庆,400065;重庆邮电大学自动化学院,重庆,400065 |
| |
| 基金项目： | 重庆市科委重大科技攻关项目，电子信息产业化发展基金 |
| |
| 摘 要： | 针对谱减法增强语音后残留"音乐噪声"明显,影响语音清晰度的问题,在分析现有的噪声谱减算法基本特性的基础上,提出一种基于噪声短时谱动态估计的语音降噪方法.通过估计噪声短时功率谱及其变化趋势,动态调整谱减法中的过减系数,有效抑制残留噪声.仿真实验表明,新的谱减算法在提高语音信噪比的同时保持较小的失真度,具有较好的增强效果.
|
| 关 键 词： | 语音识别 语音增强算法 噪声短时谱估计 谱减法 过减系数 |
| 收稿时间： | 3/1/2010 12:00:00 AM |
Spectral subtraction algorithm for speech enhancement based on dynamic estimation of noise short-time spectrum |
| |
| LI Yin-guo,XUE Wen,XU Yang.Spectral subtraction algorithm for speech enhancement based on dynamic estimation of noise short-time spectrum[J].Journal of Chongqing University of Posts and Telecommunications,2010,22(2):127-130. |
| |
| Authors: | LI Yin-guo XUE Wen XU Yang |
| |
| Institution: | College of Automation, Chongqing University of Posts and Telecommunications, Chongqing 400065, P. R. China |
| |
| Abstract: | In order to reduce musical noise in spectral subtraction method and improve the intelligibility of enhanced speech, the basic characteristics of exiting spectral subtraction were analyzed. A noise-robust method based on dynamic estimation of noise short-time spectrum was proposed. Over-subtraction of spectral subtraction could be modified dynamically to suppress residual noise effectively by estimating noise short-time power spectrum and its trends. Experimental results show that the new spectral subtraction algorithm improves the speech signal to noise ratio while maintaining a smaller distortion and it has good enhancement effect. |
| |
| Keywords: | speech recognition speech enhancement algorithm noise short-time spectrum estimation spectral subtraction method over-subtraction factor |
| 本文献已被 万方数据 等数据库收录！ |
| 点击此处可从《重庆邮电大学学报(自然科学版)》浏览原始摘要信息 |
| 点击此处可从《重庆邮电大学学报(自然科学版)》下载免费的PDF全文 |
