摘要： 研究人脸五官局部特征在人脸识别中的作用，按五官所起作用大小将其量化，量化后的值即为作用权值。对五官定位并划分区域，每个区域采用主成分分析进行特征选择。利用改进遗传算法计算每个区域特征向量的数量，将其组合使识别率达到最大。实验结果表明，利用支持向量机进行识别，在识别率最大的前提下，各个区域特征向量所占比例即为作用权值。

关键词: 人脸识别, 遗传算法, 面部区域划分, 特征提取, 作用权值, 支持向量机

Abstract: This paper researches the effect of five senses’ local features in face recognition, according to the effect of five senses, these features are quantized. The value quantized is called effect weight. Five senses are located and divided into five regions, each region uses Principal Component Analysis(PCA) to selecting features. The improved genetic algorithm is used to get the number of each region’s features while maximizing the recognition rate. Experimental results show the proportion of each region is the effect weight in this situation when Support Vector Machines(SVM) is used to recognition.

Key words: face recognition, genetic algorithm, face region partition, feature extraction, effect weights, Support Vector Machines(SVM)

中图分类号: 

• TP391