Title:Ear Identification by Fusion of Segmented Slice Regions using Invariant Features: An Experimental Manifold with Dual Fusion Approach

Abstract:This paper proposes a robust ear identification system which is developed by fusing SIFT features of color segmented slice regions of an ear. The proposed ear identification method makes use of Gaussian mixture model (GMM) to build ear model with mixture of Gaussian using vector quantization algorithm and K-L divergence is applied to the GMM framework for recording the color similarity in the specified ranges by comparing color similarity between a pair of reference ear and probe ear. SIFT features are then detected and extracted from each color slice region as a part of invariant feature extraction. The extracted keypoints are then fused separately by the two fusion approaches, namely concatenation and the Dempster-Shafer theory. Finally, the fusion approaches generate two independent augmented feature vectors which are used for identification of individuals separately. The proposed identification technique is tested on IIT Kanpur ear database of 400 individuals and is found to achieve 98.25% accuracy for identification while top 5 matched criteria is set for each subject.