摘要： 介绍了独立成分分析在全局人脸识别中的应用,重点研究了全局人脸的独立元表示,给出了基于权向量幅值、基于比例因子和基于PCA-ICA算法的3种IC获得途径。基于Umist人脸库的实验结果表明,IC表示识别率明显高于PC表示,其中基于比例因子的IC表示识别效果最好。

关键词: 独立成分分析, 主成分分析, 人脸表示

Abstract: This paper introduces independent component analysis and its application in global face recognition. Independent component kernel facial representation is focused and IC subset is obtained in three methods. Test on Umist lab image set shows that IC presentations have higher accuracy than PC representation, and the scale factor method is a most encouraging one.

中图分类号: 

• TP391.4