摘要
虽然子模式局部保持映射算法(Sub-pattern Locality Preserving Projections,SpLPP)对外部因素如光照、表情等变化具有较好的鲁棒性,但是流形的LPP在人脸识别过程中常常碰到奇异值问题,所以提出了一种利用奇异值分解的SpLPP方法,以此解决奇异值问题。该算法的关键点是将样本数据映射到一个非奇异正交矩阵中,然后再根据SpLPP求出新样本空间的低维投影子空间。在标准人脸数据库(ORL、YALE)上进行验证,实验结果表明改进的子模式局部保持映射算法在人脸识别中的有效性。
Although Sub-pattern lcality Preserving Projections is robust to variation in illumination, expression and so on, but based manifold LPP is known to suffer from singular value problem, so a solution scheme using singular value decomposition was proposed for SpLPP. The important problem of this algorithm is that the sample data were projected on a non-singular orthogonal matrix, then the data of the low dimensional sample space projection subspaee were obtained according to the SpLPP method. The experimental results on the standard face databases （ORL, YALE ） demonstrate the efficacy of the improved SpLPP approach for face recognition.
出处
《信息技术》
2013年第9期42-45,共4页
Information Technology
基金
宝鸡文理学院院级重点项目(ZK10168)
关键词
子模式局部保持映射
奇异值分解
人脸识别
Sub-pattern Locality Preserving Projection
Singular Value Decomposition
face recognition
