基于模极大值和尺度理论的音频降噪方法

朱俊敏;张潇;王旌阳;吴粤北

振动与冲击 ›› 2009, Vol. 28 ›› Issue (11) : 168-172.

PDF(1392 KB)
PDF(1392 KB)
振动与冲击 ›› 2009, Vol. 28 ›› Issue (11) : 168-172.
论文

基于模极大值和尺度理论的音频降噪方法

  • 朱俊敏1;张潇2;王旌阳1;吴粤北2

作者信息 +

AUDIO DENOISING METHOD BASED ON WAVELET TRANSFORM MODULUS MAXIMUM AND SCALE TRADING THEORY


  • Junmin Zhu1;Xiao Zhang2;Jingyang Wang1;Yuebei Wu2
Author information +
文章历史 +

摘要

在数字化时代,音频的转录或录制都会引入噪音,但是历史音频保存和音频资料处理需要纯净的音频信号,因此音频降噪研究有着重要的现实意义。该文首先介绍了二进小波和奇异性指数,并阐述了尺度跟踪和模极大值重构等理论,在Mallat工作的基础上,提出了一种基于小波滤波的音频降噪方法。该方法首先引入补偿因子削减二进小波变换对系数造成的影响,并计算带噪音频的小波系数和模极大值;然后基于信号和噪声奇异指数不同的特点,结合阈值降噪和尺度跟踪理论,采用层间相关搜索去除噪声的模极大值;最后利用交替投影算法,重建音频信号。本文用该方法处理带click和hiss噪声的音频信号,跟小波阈值方法和小波包方法相比,能达到较好的听觉效果和信噪比。同时观察信号的波形图及模极大值演示图,发现本方法都表现出优异的降噪效果。

Abstract

Audio recording or transcription inevitably brings in noise, so audio denoising is crucial for data processing and preservation,where are many existing techniques based on the wavelet transformation. Here we provide an improved algorithm based on three techniques as dyadic wavelet, scale trading theory and modulus maximum theory. The novelity of our algorithm is the following: Firstly, we introduce the compensation factors in order to reduce the influence of scale discretization; Secondly, we use interlayer correlation searching to eliminate noise according to the modulus maximum; Finally, we reconstruct the original signal using alternating projection algorithm. As an experiment, we use our algorithm to process audio signals with click and hiss noise, and achieve better results compared to the wavelet threshold and wavelet packet algorithms in terms of comfort of hearing sense. At the same time, measured by signal oscillograms and modulus maxima, this algorithm reduces the noise effectively.

关键词

小波变换 / 模极大值 / 奇异性指数 / 尺度跟踪 / 阈值降噪 / 音频降噪

Key words

wavelet transformation / modulus maxima / singularity index / scale trading / threshold / audio denoising

引用本文

导出引用
朱俊敏;张潇;王旌阳;吴粤北. 基于模极大值和尺度理论的音频降噪方法[J]. 振动与冲击, 2009, 28(11): 168-172
Junmin Zhu;Xiao Zhang;Jingyang Wang;Yuebei Wu. AUDIO DENOISING METHOD BASED ON WAVELET TRANSFORM MODULUS MAXIMUM AND SCALE TRADING THEORY[J]. Journal of Vibration and Shock, 2009, 28(11): 168-172

PDF(1392 KB)

Accesses

Citation

Detail

段落导航
相关文章

/