The Research Of Face Recognition Technology Based On Support Vector Machine & Multilevel B-splines

This paper merges theory with experiments to explore the two crucial segments of human face recognition system, known as feature extraction and classifier design. After lots of reference to literature about face recognition, this article adopts a new method--Face Recognition based on Multilevel B-Splines and Support Vector Machine (SVM). Like height in the real 3-D world, the gray value of face image is seen as a piece of height field virtually. The height of each pixel represents its gray value. Then multilevel B-Splines method is employed to approximate this height field and yield control matrix of approximation, which serves as feature vector for recognition. The following work is the construction of face image classifier using support vector machine. According to the perspective of design, the fist step is the detailed analysis of elementary principle, followed by the choice of punish coefficient,kernel function, classification method and so on. The several key parameters of support vector machine determine the overall quality of classification. The vector acquired by Multilevel B-Splines is input into the Support Vector Machine to realize the recognition and classification of face image. At last the experiment on ORL face database proves the validity of the above method. The results show satisfactory velocity and recognition rate, verifying the superiority of the recognition method based on Multilevel B-Splines and Support Vector Machine presented in this article.