摘要
基于奇异值向量方法进行人脸识别时,由于提取的奇异值向量特征所包含的人脸图像的有效信息少,导致人脸识别率低下。基于此提出了一种基于奇异值分解的人脸识别新方法——矩阵的秩-逼近法。利用ORL人脸数据库进行实验,并采用最近邻决策规则来进行分类识别。实验结果显示,提出的方法比基于奇异值人脸识别方法具有优越性,本算法能大大地改善识别效果。
When doing face recognition based on the singular value vector,the reason of the low rate for face recoginition was that the feature of singular value vector could not contain enough face information.And then,a new method of face recognition is proposed based on singular value decomposition-the method of approximation for matrix whose rank is one.Experimented on the ORL,face database,and also,the nearest neighbor decision rule is used to recgnize human face.The results of experiment reveal that this method is superior to the method of the singular value vector.Besides,this algorithm can greatly improve the effect of face recgnition.
出处
《科学技术与工程》
2010年第25期6204-6208,共5页
Science Technology and Engineering
关键词
人脸识别
奇异值分解
秩一矩阵
face recognition singular value decomposition the matrix whose rank is one
