改进核密度估计确定最优分组方法研究

摘要

Optimal grouping is an important problem of histogram algorithm to be resolved,and how to determine the group number has not a quantitative rule yet.So,we adopt the improved kernel density estimation algorithm to get the probability density function (PDF) of parameter,then define the nearness degree between the upper contour line of histogram and the PDF of parameter and use this degree to measure the closeness between them and as the judgment rule of optimal grouping,The improved kernel density estimation algorithm can get the bandwidth approaching the theoretical optimal one.We use the improved kernel density estimation algorithm to determine the optimal grouping and apply it in parameter analysis on radar emitter source signal,the result indicates that this method is effective and can search the optimal grouping automatically.%最优分组问题是直方图算法需要解决的一个重要问题,对于分组数如何确定没有一个定量的规则.为此,采用改进的核密度估计算法得到参数的概率密度函数,然后定义直方图上部轮廓线与参数概率密度函数之间的贴近度,以此度量直方图上部轮廓线与参数概率密度函数之间的接近程度,作为最优分组的判决准则.改进的核密度估计算法可以获得接近于理论最优窗宽,利用改进核密度估计算法确定最优分组并用于雷达辐射源信号的参数分析中,结果表明该算法是有效的,可以自动搜索出最优分组数.