Research On Identity Validation System Based On Face Recognition

Biometric technique, which makes use of physical and behavior characteristics of human for personal identification, has got more attention among many personal identification methods, for its higher security, reliability and effectiveness. Human face recognition is a biometric technique that is easily accepted. It has important research and application values and has been an active research in recent years.In this paper, we study and implement an identity validation system based on face recognition, this paper contains four parts as follow.(1) Face skin color has a relatively convergent distribution in color space. Face can be extracted from the context making use of this characteristic. Then feature extraction and validation of face can be done to detect face domain.(2) the objective of which is to prepare for face characteristic extraction. The main methods of it include format transformation, noise cancellation, edge detection, binarization, and normalization of face image.(3) Fourier transformation and singular value decomposition methods are adopted to extract algebraic features of face images. First, face image is Fourier transformed to get its the amplitude representation. Second, the amplitude representation is projected onto two compressed orthogonal matrixes, which come from the SVD of the standard face image obtained by averaging all training samples. The projecting coefficients can be used as the algebraic feature of the face image.(4) Last, an improved Nearest Neighbor Classifier (NNC) is adopted to recognize the unknown face by using Euclidean distance as the similarity measurement.A whole personal identification system is designed in this paper, which combines algorithm optimization, system design. The system takes many factors into account, such as image requisition devices, scene setting, etc. to exert the algorithm efficiency.