摘要
为消除乘法性观测噪声,利用独立分量分析的冗余取消特性,提出一种基于同态变换盲源分离(BSS)的消噪新方法.通过对原始观测信号的同态变换———对数变换,将乘法性噪声转变为加法性噪声.引入经过相同对数变换的虚拟噪声分量,将变换后的一维观测信号扩展为多维观测.依据最大化估计分量间独立性的准则(由高阶累积量构成)对多维观测实施盲源分离.最后,联合指数逆变换及输出信号校正处理,实现真实信号的估计,从而消除原始观测中的乘法性噪声.仿真消噪实验结果表明,该方法可有效地消除观测信号中的乘法性噪声.与同态滤波等传统的乘性噪声消除技术相比,新方法不仅实现简单,运算效率高,而且可以方便地扩展应用于多维观测信号的噪声消除.
In order to remove multiplicative noise in observation a new method for multiplicative noise reduction based on homomorphic transform and blind source separation （BSS） was proposed, using redundancy reduction of independent component analysis （ICA）. Original observation with multiplicative noise was first transformed by the new method into one with additive noise by means of homomorphic transform, i.e. a logistic operation. And, virtually noisy component was incorporated into the transformed observation, so that a multi-dimensional observation vector was constructed. Then, blind source separation （BSS） was applied to extract independent components from the extended observation vector, according to principle of maximizing statistical independence between estimated components, which was formulated by cumulates higher than second order. Finally, the subsequent processing combining an exponential transform with an output revise was used to restore the true signal. Thus, the multiplicative noise embedded in observation was removed. Experiment with synthetic vibration data verified that the proposed method could remove multiplicative component in noisy observation effectively. Compared with some traditional techniques such as homomorphic filter etc. , it is not only simple and efficient but also very easy to be extended and used for noise reduction in multi dimensional observation.
出处
《浙江大学学报（工学版）》
EI
CAS
CSCD
北大核心
2006年第4期581-584,614,共5页
Journal of Zhejiang University：Engineering Science
基金
国家自然科学基金资助项目(50505016
50205025)
浙江省自然科学基金资助项目(Y105083)
关键词
独立分量分析
盲源分离
冗余取消
同态变换
乘性噪声消除
independent component analysis （ICA）
blind source separation （BSS）
redundancy reduction
homomorphic transform
multiplicative noise reduction
