摘要
提出了一种基于奇异值分解和数据融合的脸像鉴别方法 .该方法首先利用奇异值分解方法 ,求出脸像矩阵的奇异值及奇异值向量 ,分别利用所求得的奇异值及奇异值向量作为特征矢量进行脸像鉴别 ,分别得出基于奇异值和重建误差的鉴别结果 ,此结果以隶属度函数方式表示 .将上述鉴别结果用 L OGISTIC回归方法进行融合 ,得出更为准确的脸像鉴别结果 .该方法克服了“小样本”效应并引入正负样本学习过程 ,提高了正确鉴别率 .利用 ORL人脸数据库进行实验 。
A face identification method based on singular value decomposition (SVD) and data fusion is proposed in this paper. The singular values (SVs) and singular value vectors of a face image matrix are extracted and employed as features. These features are matched using SVs and singular value vectors respectively. The matching outputs are based on SVs and reconstruction errors and expressed by membership function values. For more accurate results, these membership function values are fused by LOGISTIC regression. The proposed identification technique improves the correct verification rates from the following reasons. First, fusion makes accurate identification results. Second, the method solves the problem of small sample size that is difficult to avoid in face recognition problem. Third, the LOGISTIC regression fusion method has the learning ability. So it can learn both “positive” and “negative” samples and the correct identification rates are achieved. The ORL face database is used in experiment and experiment results show that the novel face verification method is effective and possesses several desirable properties when it compared with many existing methods.
出处
《计算机学报》
EI
CSCD
北大核心
2000年第6期649-653,共5页
Chinese Journal of Computers
基金
国家"八六三"高技术研究发展计划!( 863 -3 17-0 1-10 -99
863 -3 0 6-ZT-0 60 0 6-5 )
百人计划和中国博士后基金
关键词
脸像鉴别
奇异值分解
数据融合
图像识别
face identification, singular value decomposition(SVD), data fusion
