一种噪声谱快速跟踪的语音增强方法

山东大学学报(工学版)

一种噪声谱快速跟踪的语音增强方法

陈文钢 ^1,2，田岚²，姜晓庆²，孙英明^*

1. 山东理工大学电气与电子工程学院，山东淄博255049；2. 山东大学信息科学与工程学院，山东济南250100

收稿日期:2006-03-20 修回日期:1900-01-01 出版日期:2006-08-24 发布日期:2006-08-24
通讯作者: 陈文钢^1,2

A speech enhancement method based on quickly tracking the noise spectrum

CHEN Wengang^1,2，TIAN Lan²，JIANG Xiaoqing²，SUN Yingming^*

1. School of Electrical and Electronic Engineering,Shandong University of Technology,Zibo 255049,China；2. School of Information Science and Engineering,Shandong University,Jinan 250100,China

Received:2006-03-20 Revised:1900-01-01 Online:2006-08-24 Published:2006-08-24
Contact: CHEN Wengang^1,2

摘要/Abstract

摘要： 提出了基于多窗谱法和小波阈值法对带噪语音的短时功率谱进行估计的语音增强改进方法.针对软、硬阈值法自身的缺点，采用了一种改进方法，即模平方处理方法.噪声估计过程中通过跟踪带噪语音帧来更新噪声估计.仿真结果表明，本文提出的方法能够快速跟踪突变噪声，具有较强的降噪能力.把此算法综合到语音增强中时，其增强性能要好于STSAMMSE算法.

关键词: 语音增强, 多窗谱法, 小波阈值, 噪声估计

Abstract: A new speech enhancement method based on multitaper spectrum method and wavelet thresholding is proposed so that a noisy speech power spectrum is obtained. As for the faults of the hardthresholding and softthresholding, an improved model of threshold (the modulus squared mode1) is presented. During the noise estimation, the estimation of its spectrum is updated by tracking the speechabsent frames. The experimental results show that this method presented here can quickly track the changes of the speech and has better denoise function. When combining the method with speech enhancement, the enhancement result is better than that of STSA algorithm.

Key words: multitaper spectrum method, wavelet thresholding, noise estimation , speech enhancement

陈文钢 ,田岚,姜晓庆,孙英明 . 一种噪声谱快速跟踪的语音增强方法[J]. 山东大学学报(工学版), 2006, 36(4): 26-28 .

CHEN Wengang,TIAN Lan,JIANG Xiaoqing,SUN Yingming*

参考文献

多维度评价

Viewed

Full text

Abstract

Cited

Shared

Discussed