Dual Gradient Feature Pair Based Face Recognition for Aging and Pose Changes

Abstract

This paper proposes a novel dual gradient feature pair (DGFP) based face recognition system that is robust against pose and age variation. This DGFP method, initially segments the rectangular face region into 6 regions which forms 3 pairs which is almost symmetrical to each other. From each pair highly matched HOG features are extracted and the dual gradient features like the positive gradients and negative gradients are extracted from the symmetrical regions of the face gets matched. DGFP is extracted in both the training and testing images. The matching is obtained using the minimum Euclidean distance with a threshold value. Experimental results are evaluated using the datasets such as MORPH, FGNET and Yale dataset and the results reveals that the proposed DGFP face recognition system has achieved an average recognition rate of 92.07%, average time of feature extraction is 253s and the average time of face recognition is 0.1533s