摘要
信号检测和分析经常受到毛刺的干扰,因此有必要将其去除.小波变换作为一种理想的时频分析方法,能够对信号进行多尺度或多分辨率分析,从不同层次把握信号的特征信息,为此提出了一种基于小波变换模极大值与多重分形的信号毛刺去除方法,分别利用多重分形谱、局部有效Hlder指数实现毛刺的检测与定位,进而实现毛刺的去除.实验证明,这种方法是可行和有效的,为非平稳随机信号中的毛刺去除提供了一种行之有效的方法.
Outliers often interfere with signal detection and during analyzing processes and must therefore be removed.As an ideal time-frequency analysis method the wavelet transform provides a multiscale（or multiresolution）signal analysis and captures the characteristic information at different levels.An outlier removal method based on WTMM and multifractal was thus presented,which utilized multifractal spectrum and the local effective Hlder exponent to detect and localize outliers respectively,so as to eventually remove then.Experimental results show that the proposed method is feasible and valid,and additionally,it provides an effective method for outlier removal in nonstationary stochastic signals.
基金
国家自然科学基金(60872163)资助
