一种通过变量分配改进脸谱图的思路与方法研究

发布时间：2018-04-15 04:00

本文选题：Chernoff脸谱图 + 方差　；参考：《调研世界》2017年09期

【摘要】：针对现有Chernoff脸谱图(以下简称脸谱图)的绘制过程并未考虑变量分配对其最终表达能力的影响,本文尝试提出改进脸谱图的思路。首先通过人类对不同脸谱图物件(包括脸部器官和发型等)的识别敏感度进行降序排序;其次对变量的变异程度依方差、变异系数、极差同样进行降序排序;再次依强强结合的原则将变异程度高的变量分配到识别敏感度高的(变异程度低的变量则分配到识别敏感度低的)构件上,实现脸谱图的改进,增强脸谱图的表达能力;最后利用人体尺寸数据与树叶轮廓数据对几个方案进行比较,得出基于方差与极差的变量分配所得到的脸谱图明显优于现行基于变量随机分配的脸谱图的结论。
[Abstract]:In view of the fact that the mapping process of Chernoff (hereinafter referred to as Facebook) does not take into account the effect of variable allocation on its ultimate expressive ability, this paper attempts to put forward the idea of improving the facial graph.First, the human beings rank the sensitivity of different facial images (including facial organs and hairstyles) in descending order; secondly, the degree of variation of variables is ranked according to variance, coefficient of variation and range.Thirdly, according to the principle of strong and strong combination, the variable with high degree of variation is assigned to the component with high sensitivity (the variable with low degree of variation is assigned to the component with low sensitivity of identification), so that the improvement of facial graph is realized and the expression ability of facial graph is enhanced.Finally, several schemes are compared by using human dimension data and leaf contour data. It is concluded that the face map based on variance and range assignment is obviously superior to the existing facial map based on random assignment of variables.
【作者单位】：中国人民大学统计学院;
【分类号】：C81

【相似文献】