基于稀疏表示的人脸识别系统设计与实现  

The design and implementation of face recognition system based on sparse representation

作　　者：廖延娜[1] 马超[2] 

机构地区：[1]西安邮电大学理学院,陕西西安710121 [2]西安邮电大学电子工程学院,陕西西安710121

出　　处：《电子设计工程》2016年第17期153-155,159,共4页Electronic Design Engineering

基　　金：陕西省教育厅专项科研计划项目(12JK0559);陕西省自然科学基础研究计划项目(2014JM2-6117)

摘　　要：针对人脸识别在实际应用中存在光照异常、姿态变化、遮挡、样本缺乏等情况,研究了结合主成分分析的稀疏表示人脸识别分类算法,设计并实现了人脸识别系统。系统依据训练字典对实时采集的人脸图像进行识别,并支持训练字典的实时更新。测试结果表明,在训练样本充分的情况下,系统对正面无表情人脸识别率在99%以上,满足正常人脸识别的使用需求;在训练样本不充分的情况下,识别率仍可以达到70%以上,可用于安防预警。In order to solve the situations of abnormal light, attitude change, obscured, samples lack in actual application of human face recognition system, the recognition algorithm based on sparse representation combined with principal component analysis is studied. A human face recognition system is designed to realize real time face image recognition in real time according to training dictionary, and the training dictionary can be updated real time. According to the test results, when the training samples are adequate end the face images are expressionless, frontal, the rate of system identification over 99%,meetting the demand of normal use of face recognition. when the training samples are not adequate, the rate over 70%, can be used in the security warning.

关 键 词：计算机图像处理 人脸识别 稀疏表示 主成分分析 

分 类 号：TN911.73[电子电信—通信与信息系统]