Support vector machines for face recognition

Face recognition is one of the challenging problems in human-computer interaction. An automated face recognition system requires an efficient method for detection of face region in the image sequence, extraction of facial features, and construction of a recognition model. In recent years, support vector machines (SVMs) have demonstrated excellent performance in a variety of pattern recognition problems. In this paper, we apply SVMs for face recognition (recognition model). Multi-class recognition system using SVMs are built using onc-against-the-rest approach. In this approach, one SVM model is built for each persons face (class). For each SVM model an optimal hyperplane is constructed in the kernel feature space to separate the examples of a class from the examples of all the other classes. SVMs learn the boundary regions between patterns of two classes by mapping the patterns into a higher dimensional space, and seeking a separating hyperplane, so as to maximize its distance from the closest training examples. SVM based approach for face recognition has been demonstrated for partial CMU face data base. The face recognition system is evaluated on faces of 10 different persons. For each person there arc 75 faces. The results of our studies show that, the system gives about 100% detection rate.