计算机工程与应用 ›› 2011, Vol. 47 ›› Issue (24): 143-145.

• 图形、图像、模式识别 • 上一篇    下一篇

基于状态估计的张量分解人脸识别方法

梅蓉蓉,吴小俊,冯振华   

  1. 江南大学 信息工程学院,江苏 无锡 214122
  • 收稿日期:1900-01-01 修回日期:1900-01-01 出版日期:2011-08-21 发布日期:2011-08-21

Face recognition method based on state estimation and tensorfaces algorithm

MEI Rongrong,WU Xiaojun,FENG Zhenhua   

  1. School of Information Engineering,Jiangnan University,Wuxi,Jiangsu 214122,China
  • Received:1900-01-01 Revised:1900-01-01 Online:2011-08-21 Published:2011-08-21

摘要: 张量脸算法是分析和表达多因素影响的人脸图像结构的一种有效的数学模型,然而张量分解对状态空间的非线性处理仍存在不足之处。对此提出了一种新的多姿态人脸图像识别方法,在原有的张量脸算法基础上结合状态估计的方法。将训练样本图库中不同状态的人脸通过PCA分解得到多种状态(角度、光照、表情)分别对应的特征空间,对于测试样本先投影到每个特征空间,利用最近邻分类器进行状态估计,对利用张量脸算法得到的张量脸进行识别。实验结果表明,该特征提取方法的识别率优于原有的张量脸算法。

关键词: 张量脸, 状态估计, 人脸识别, 最近邻分类器

Abstract: Tensorfaces algorithm is an effective mathematical model which can analyze and express the frames of multi-view face images,but there are some problems of multi-linear analysis method with nonlinear changes of face images.So an improved tensorfaces algorithm is proposed for multi-view face recognition which integrates state estimation.The training face images from different states are decomposed to some eigenspaces(views,illuminations and expressions) by PCA.Then the testing face images can be projected into each eigenspace and estimate the states of the unknown images by the closest classifier.It can recognize the faces by tensorface of every image which is obtained by the tensorfaces algorithm.Experimental results show that this method outperforms the original tensorfaces method.

Key words: tensorfaces, state estimation, face recognition, closest classifier