基于Gabor特征和EHMM的人脸识别方法  

Face recognition based on Gabor feature and EHMM

作　　者：杨勇[1] 田侃[1] 

机构地区：[1]重庆邮电大学计算机科学与技术研究所,重庆400065

出　　处：《重庆邮电大学学报（自然科学版）》2011年第3期355-362,共8页Journal of Chongqing University of Posts and Telecommunications(Natural Science Edition)

基　　金：重庆市杰出青年基金(2008BA2041);重庆邮电大学基金(A2009-26);重庆市计算机网络与通信技术重点实验室开放基金(CY-CNCL-2009-02)~~

摘　　要：针对传统人脸识别方法在受到表情、姿态和光照等影响时存在鲁棒性差的问题,提出了一种改进的人脸识别方法。在特征提取上,该方法首先提取人脸的Gabor特征,接着利用DCT(discrete cosine transform)压缩使得相似特征得到聚合,最后利用PCA(principal component analysis)分别筛选出最能够代表人脸各个区域的DCT系数。在识别方法上,该方法采用了嵌入式隐马尔科夫模型(embedded hidden markov model,EHMM),并基于人脸的认知结构信息对嵌入式隐马尔可夫人脸结构模型进行了改进。对比实验结果表明,该方法识别率高,复杂度低,并具有较好的鲁棒性,易于在工程上应用。This paper brings forward a new face recognition method,aiming at improving the inefficient robustness of the traditional face recognition method when factors like expression,gesture,and illumination are involved.When it comes to feature extraction,this new method will first extract the Gabor feature of a human face,and then assemble the similar features with DCT（discrete cosine transform）,and last,sieve out the DCT coefficients which can mostly represent each areas of human face with PCA（principal component analysis）.When it comes to recognition method,it adapts EHMM（embedded hidden markov model）,and improves the EHMM human face model according to the understanding of cognitive structure of human face.Comparative experiments show that the proposed method has a high recognition rate and a low complex rate.And it has excellent robustness,which proves that it can be easily applied to engineering projects.

关 键 词：人脸识别 嵌入式隐马尔科夫模型 GABOR小波变换 离散余弦变换 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]