基于正交SRDA和SRKDA的人脸识别  

Face recognition based on orthogonal SRDA and SRKDA

作　　者：陈达遥[1] 陈秀宏[1] 董昌剑[1] 

机构地区：[1]江南大学数字媒体学院,江苏无锡214122

出　　处：《计算机应用研究》2014年第1期299-303,320,共6页Application Research of Computers

基　　金：中央高校基本科研业务费专项资金资助项目(JUSRP211A70);国家自然科学基金资助项目(61373055)

摘　　要：利用正交投影技术进行降维可以更好地保留与度量结构有关的信息,提高人脸识别性能。在谱回归判别分析(SRDA)和谱回归核判别分析(SRKDA)的基础上,提出正交SRDA(OSRDA)和正交SRKDA(OSRKDA)降维算法。首先,给出基于Cholesky分解求解正交鉴别矢量集的方法,然后,通过该方法对SRDA和SRKDA投影向量作正交化处理。其简单、容易实现而且克服了迭代计算正交鉴别矢量集的方法不适应于谱回归(SR)降维的缺点。ORL、Yale和PIE库上的实验结果表明了算法的有效性和效率,在有效降维的同时能进一步提高鉴别能力。The dimensionality reduction by orthogonal projection techniques helped preserve the information related to the metric structure and improved the recognition performance in face recognition. Based on spectral regression discriminant analysis （SRDA） and spectral regression kernel discriminant analysis （SRKDA）, this paper proposed two dimensionality reduction algorithms named orthogonal SRDA （OSRDA） and orthogonal OSRKDA （OSRKDA）. Firstly, it gave a set of orthogonal discriminant vectors obtained based on Cholesky decomposition. Then, this paper orthogonalized the projection vectors of SRDA and SRKDA by this method. It was very simple and easy to implement. What＇s more, it overcame the shortcoming that the iterative algorithm of orthogonal discriminant vectors was not suitable for spectral regression dimensionality reduction algorithms. Experiments on ORL, Yale and PIE demonstrate the effectiveness and efficiency of the algorithms, and show that these algorithms can reduce the dimensions of the data and improve the discriminant ability.

关 键 词：降维 人脸识别 谱回归 正交鉴别矢量 CHOLESKY分解 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]