Design and Implementation of Support Vector Based Classifier for Face Recognition

Linear Discriminant Analysis (LDA) is a well known technique for face recognition with good recognition rate. Support Vector Machines (SVMs) have been recently proposed as a new classifier for pattern recognition. A novel method for face recognition, using LDA and SVM has been proposed in this paper wherein large sample size problem is reduced to small sample size problem using support vectors, and Discriminant analysis is done only on support vectors. Locality preserving projection (LPP) is used as a basic dimension reduction technique. Experiments are performed on Indian Face database and error rates of classification and elapsed time for performance evaluation are compared with techniques such as locality preserving projection LPP , LPP and principal component analysis (LPP+PCA) and LPP and linear discriminant analysis(LPP+LDA).