Application Of Linear Projection Analysis In Face Recognition

Face recognition is one of biometric methods identifying individuals by the features of face. Research in this area has been conducted for more than 30 years. As a result, the current status of face recognition technology is well advanced. How to extract the available feature is the key to the problem of face identification. Linear projection analysis is one of the classical and widely used techniques for feature extraction.Some of their theories and algorithms are researched further in this dissertation, and present the ICA+LDA algorithm, which combined with the advantages of Independent Component Analysis (ICA) and linear discriminant analysis (LDA). Feature sub-space of training samples is obtained by way of ICA, and feature sub-space from LDA is calculated on the basis of ICA. In the meanwhile, the two feature sub-spaces from ICA and LDA are fused, and the fusion feature space is acquired. After training samples and test samples are respectively projected towards the fusion feature space, recognition features are accordingly gained. Nearest neighbor rule is utilized in gender classification. Face recognition software was developed using the algorithm of ICA+LDA based on ORL (Olivetti Research Laboratory) face database, and improved the algorithm.Another topic to be discussed here is the design and implementation of the face recognition test system. The system offers an environment of a uniformed poll for achieving and testing face recognition algorithms. It supplies for the further researching of parameter with experiment tools and algorithm integrations. When planning a certain method, we keep to the interface definitions of the system, designed for application integration. We realize the algorithms in the form of module objects, following the interface. All work related to data I/O, displaying, analyzing are given to the system. The system also provides us with the ability of increment development and secondary development. It is convenient for us to extend datum by video capture of camera on our demands. All function models are designed, generated and added into the system under the idea of software components.