基于SVD和KL投影相融合的人脸识别算法

Face recognition based on fusion feature extraction from SVD and KL projection

在已有的PCA人脸识别算法的基础上构建了一种基于奇异值分解和KL投影相融合的人脸识别方法。方法对人脸图像进行奇异值分解与KL变换,随后将KL投影的主特征向量和SVD特征向量进行线性融合形成新的特征向量,并以此作为判别标准。经试验测试该方法消除了当检测库中人脸的姿态变化时人脸图像间的相关性下降对识别正确率的影响,大大的提高了人脸识别的正确率。

A new approach based on the fusion feature extraction from the singular value decomposition （SVD） and the KL projection is presented on the basis of the PCA face recognition method. This method carries on the SVD and the KL transformation in the face image. The main characteristic,vector of the KL projection and the SVD characteristic vector is fused to form a new one which is taken as the distinction standard. The experiment result shows that this method eliminates the recognition influence by the relevance drop between the face image when the posture changes in the examination storehouse and improves the recognition rate significantly.