Ear recognition using pyramid histogram of orientation gradients
Partha Pratim Sarangi, Bhabani Shankar Prasad Mishra, Satchidananda Dehuri · 2017
Ear recognition Is still a standing problem In biometrics and has become an open research area in recent years. In this paper, we explore a new local feature extraction technique pyramid histogram of oriented gradients (PHOG) to represent ear images. However, the PHOG descriptor of the ear image is significantly large. To reduce the dimension of the PHOG descriptor, linear discriminant analysis (LDA) has been used to remove noise and avoid over fitting. Finally, the discriminant features are classified using nearest neighbor classifier. The PHOG has inherent properties to efficiently handle the problems of change in pose and partial occlusion of the ear images. The results of the proposed method are evaluated using two public ear databases, namely IIT Delhi ear database and University of Notre Dame ear database (Collections E). Our method with reduced features using LDA offers promising results and largely improves the recognition accuracy over existing methods.