Face and Iris biometrics person identification using hybrid fusion at feature and score-level

Face and Iris biometrics are amongst the studied unimodal systems by researchers over the past years due to their ease of acquisition and accuracy, respectively, during the recognition process. However unimodal systems are not perfect when deployed to real-world applications due to non-ideal conditions such as off-angle gaze, illumination, occlusion and variation in posing. These limitations have led to an increase of research in multi-biometrics. In recent times, researchers have worked on combining unimodal templates with different methods with each having to compensate for the shortcomings of the unimodal systems. In this work we present a hybridized fusion strategy that combines, three classifiers based on feature and score level fusion using a decision level fusion rule. We compare the recognition rate of the proposed method with other fusion methods in literature. We have obtained a recognition accuracy of 98.75%. The proposed method was validated using the ORL face and CASIA iris datasets.