非线性结构保持能力的不足是正则正交化的线性判别分析ROLDA(Regularized Orthogonal Linear Discriminant Analysis)在人脸识别中的主要问题。提出一个用于人脸识别的正则正交化的局部Fisher判别分析ROLFDA(Regularized Orthogonal LocalFisher Discriminant Analysis)降维算法。该算法在ROLDA基础上引入局部结构保持,继承ROLDA的特性,克服了ROLDA的非线性能力的不足的问题。在YaleB和AR人脸数据集上的实验验证了该算法的有效性。
Insufficiency of nonlinear structure preserving ability is the main problem of regularised orthogonal linear discriminant analysis(ROLDA) to be applied in face recognition.In the paper a dimensionality reduction algorithm called regularised orthogonal local Fisher discriminant analysis(ROLFDA) for face recognition is proposed.On the basis of ROLDA,the algorithm introduces local structure preserving,which inherits the properties of ROLDA and overcomes the problem of the insufficiency of nonlinear ability.Experiments on YaleB and AR demonstrate the effectiveness of our proposed algorithm.