The Classification of Scores from Multi-classifiers for Face Verification

Abstract

We have proposed a multiple classifier systems for face verification by the study of classification of scores of the four face authentication systems built by facial feature extraction phase is filtered using Gabor wavelets and Principal Component Analysis (PCA ) plus the Enhanced Fisher linear discriminant Model (EFM) as a method of reducing data space. For the study of classification of scores we used three methods: statistical method of Fisher, the Support Vector Machine (SVM) and artificial neural networks (MLP). Another important issue addressed in this work is the normalization of scores is proposed by the classification scores, why we try to study at this stage three methods of normalization of scores: Z-Score, quadratic-linear-quadratic (QLQ) and double sigmoid function. Copyright © 2012 IFSA.