均匀度估计的噪声源盲分离算法研究

盲信号分离技术是源输入未知时识别噪声源的一种有力手段。源的独立性常是应用盲分离算法的一个基本要求。但由于概率密度函数一般未知且估计繁琐,传统盲分离算法对源信号的独立性较难检验。为此,从信号独立性分析出发,理论上推导了随机变量的独立性和其概率分布函数的联合分布之间的关系,提出了一种独立性度量指标均匀度的估计算法,并给出了相应的盲分离算法。利用电机和海水泵的振动信号分离试验对方法进行了验证,并与现有的半熵盲分离算法进行比较,结果表明在分离效果和计算时间方面均优于现有的方法。充分说明了本文算法的有效性。 

Blind separation algorithm for noise sources via uniformity estimation

Blind separation algorithm is a powerful method on identifying noise sources without any prior knowledge of sources and paths. The independence of source signals is usually a main requirement for blind separation algorithm. However, the independence can hardly be checked up because the probability density function is unknown or difficult to estimate. For solving this problem, the relationship of independence of random variable and joint distribution of its probability distribution function is illuminated theoretically. Then, an algorithm of estimating the new independence index-uniformity is presented by a simplified estimation method, and the corresponding blind separation algorithm is described. The feasibility of uniformity and blind separation algorithm were demonstrated by the mixed signals' separation experiment of electromotor and seawater pump, and the experiment results, including the separation effects and time cost, show that the proposed method is superior to the existing semi-entropy blind separation algorithms.