Ear Recognition Using Rank Level Fusion of Classifiers Outputs

An individual's authentication plays a vital role in our daily life. In the last decade, biometric-based authentication has become more prevalent than traditional approaches like passwords and pins.   Ear recognition has gained attention in the biometric community in recent years. Researchers defined several features for the identification of a person from ear image. The features play a vital role in the success of classification models. This paper considers an ensemble of features for designing a new classification model. The features are assessed in isolation as well as through feature-level fusion.  Subsequently, a rank-level fusion for classification is introduced. The experiments are conducted on both constrained and unconstrained ear datasets. The results are promising and open up new possibilities in machine learning-based ear recognition