Subspace Characteristic Extraction Algorithm Of Human Face Recognition Based On Manifold

As a biometric technology, face recognition is very important in the fields of biometric technologies.face recognition technology involves in many fields such as Image Processing, Pattern Recognition and Computer Vision.A complete face recognition system must have two parts: feature extraction part and classifier part.The feature extraction part processed the original images, projects them from high dimensionality to lower one and extracts the information which is important for the classifier. Then, class information is obtained from the features in the classifier part. Manifold learning method is a newly proposed machine learning method, which can effectively explore the inner structure of high dimensional data. It is a trend to put the essence of manifold learning into subspace feature extraction algorithms.Based on manifold learning methods, researches on both feature extraction and classifier, analyze on defects and put forward an improved algorithm of Semi-Supervised Discriminant Analysis: SDA.New algorithms is a kind of Semi-Supervised dimensions reduection baseed on part and full. Making use of labeled and unlabeled data ,considering the inner geometry structure of data and classifier distinguish information to improve algrothm's recognition feature.