Hybrid Technique to Improve Face Recognition Using Principal Component Analysis and Singular Value Decomposition

Abstract

This paper present a hybrid technique between two of the most popular face recognition methods, Principal Component Analysis (PCA) and singular value decomposition (SVD), and attempts to offer a study for all its mathematical equations in detail and concentrate on the hybrid place between equations in order to focus on the way of processing the hybrid method. Dot product used in mathematical equations and for testing the proposed method used Olivetti Research Laboratory (ORL) data set images were used with different number of images for training set and used various number of Eigen faces and used also dissimilar number for test images and Manhattan distance was used to measure the distances between image vectors in this system, the result shows that the recognition rate using this hybrid technique is higher than the recognition rate using PCA or SVD separately and each time increase the threshold value the accuracy rate increased and conclude that when increase the threshold value and the chosen number of Eigen faces then recognition rate increased.