Application Research Of Face Recognition Based On SVDD And SVM

The biological characteristics are the inherent attributes of human beings, which have strong self-stability and individual independency. They are the ideal basis for authentication. Compared with other biological characters, the patterns of face are natural and direct; therefore the technique of face verification is widely used as the main approach in ID verification. The process of facial recognition can be divided into three stages: face detection and preprocessing of face image, feature extraction and recognition. In this work, we firstly discuss the background and then analyze the main existing face recognition algorithms. With this understanding, we present the method called Hypersphere Projection Embedding Support Vector Discriminant Analysis (HPE-SVDA) for feature dimension reduction.Then hierachical face recognition with the ability of rejection for non-target and classification for target is proposed.The content is described as below:(1) Firstly, we describe the the theory of support vector machine, nuclear function, as well as the the basic ideas of support vector data description in detail. These are prepared for the new algorithm we proposed in the next step.(2) We design and realize the face detecting based on the Adaboost algorithm. And the method of face image preprocessing such as gray information normalization, dimension normalization are designed and realized synchronously. In this way, the foundation is established for the feature extraction and recognition of face.(3) The method of hypersphere projection embedding support vector discriminant analysis for feature dimension reduction is presented. The algorithm first extracts the category information using the support vector machine, then sets the distance of the sample to the sphere-center as the projection distance and normalizes the projection axis by using the sphere radius. At last, we combine the category information and normalized projection distance, the compression features of the sample can be obtained.(4) We propose and realize the hierachical face recognition algorithm based on svdd and svm. This algorithm can reject the non-target and classify the target samples at the same time. It establishes the ultra-compact package sphere model of objective samples for the acceptance of targets and the rejection of the non-targets. And then it constructs the multi-classification support vector machine model for the classification of the targets by means of the outstanding classification performance and generalization ability of svm.(5) We mainly analyze the structure and funtions of face recognition system with rejection ability and use object-oriented design methods to designe and implement a prototype of face recognition system in attendance.