摘要
由于小波基缺乏平移不变性,传统小波及小波包去噪算法可能使信号急剧变化部分产生人为振荡现象.提出了基于平移不变的小波包去噪方法,对所分析的信号进行循环平移,利用软或硬阀值对该信号的小波包系数进行压缩,重构信号,再进行相反的循环平移,通过多次的平移—消噪—平移,平均所获得的结果,从而消除小波包基的平移依赖性.对比普通小波包去噪,该方法能有效地消除人为振荡现象,使去噪后的信号更光滑,更逼近真实信号.
Denoising algorithm based on traditional wavelet packet may produce artifacts on discontinuities of the signal. The reason is that the de-noising algorithm lacks of wavelet translation invariant. This paper proposes a denoising method based on translation invariant. The method performs the cycle-spinning for the signal to be analyzed. And the soft （hard） threshold is used to shrink the wavelet packet coefficient of the signal and reconstruct the signal. Consequently, the shift dependence of wavelet packet basis is eliminated. This method can suppress the artifacts effectively so that denoised signal is more smooth and has better approximation to original signal than traditional wavelet packet.
出处
《中南林业科技大学学报》
CAS
CSCD
北大核心
2007年第3期148-150,158,共4页
Journal of Central South University of Forestry & Technology
关键词
信号分析
平移不变小波包
去噪
signal analyzed
translation invariant wavelet packet
denoising
