An efficient feature selection scheme based on genetic algorithm for ear biometrics authentication

Human ear recognition is a new biometric technology which competes with other powerful biometrics modalities such as fingerprint, face and iris. Ear has small size, a uniform distribution color and does not need much collaboration from the user. Feature extraction is a crucial stage for biometric identification. However, the extracted features might contain redundant and irrelevant features which can lead to the problem of dimension and even to degradation of performances of biometric systems. In this paper, we present a new efficient feature selection scheme based on Genetic Algorithm for ear biometrics. The proposed approach has been tested on an ear biometrics database and compared with the full feature system, Principal Components Analysis (PCA) based approach and a combination of the proposed GA and PCA. Experimental results show that the proposed approach outperforms the full-feature based system in terms of accuracy, FRR and FAR.