基于虚拟样本扩张法的单样本人脸识别算法研究  

Virtual Sample Generating for Face Recognition from a Single Training Sample per Person

作　　者：单桂军[1,2] 

机构地区：[1]江苏科技大学,电信学院,镇江212003 [2]镇江高等专科学校,电子与信息系,镇江212003

出　　处：《科学技术与工程》2013年第14期3908-3911,3916,共5页Science Technology and Engineering

基　　金：江苏省高校实验室研究会研究课题(JS2012-2);江苏省现代教育技术研究2010年度课题(16866);镇江市科技支撑计划项目(GY2012041)资助

摘　　要：随着人脸识别技术的不断发展,单样本人脸识别已成为当今的一个热点。针对单样本人脸识别问题,提出了一种基于虚拟样本扩展的人脸识别方法,为给定的单训练样本增加虚拟图像,以增强单训练样本的分类信息,并对原样本及其虚拟样本进行特征变换,划分得到更多的子图像,利用二维主成分分析(2DPCA)实现特征抽取,一定程度上减轻了人脸的表情、姿态、光照等因素对识别效果的影响,提高了识别率。提出的方法分别在ORL及FERET两大人脸数据库上得到了验证。Usually, that are assumed there are muhiple samples per person for feature extraction in many facerecognition methods. In many practical face recognition applications such as law enhancement, e-passport, and ID card identification, however, this assumption may not hold as there is only a single sample per person. To address this problem, a novel method is proposed that generating multiple samples with sample give by using virtual sample generating method so as to adding classes of each face, partitioning them into multiple sub-patches, using 2DPCA method to complete feature extraction. Experiment results on two widely used face databases are presented to dem- onstrate the efficacy of the proposed approach.

关 键 词：人脸识别 单训练样本 虚拟样本 二维主成分分析 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]