Efficient Human Recognition System using Ear and Profile Face

This paper proposes an efficient human authentication system which makes use of ear and profile face data. SURF has been used to extract features from enhanced images and fusion has been performed at feature level and score level. The evaluation has been done on three data sets viz. IITK Data Set I acquired at Indian Institute of Technology Kanpur, collection E (UND-E) and collection J2 (UND-J2) acquired at University of Notre Dame. It has been observed that the fusion of ear and profile face improves the recognition performance considerably as compared to one when ear or face are individually used. For the fusion strategy, it is observed that the score level fusion performs better than the feature level fusion and has achieved verification accuracies of 98.02% and 96.02% with EER of 2.83% and 4.4% on UND-E and UND-J2 data sets respectively and that of 99.36% with EER of 0.85% on IITK Data Set 1.