Unconstrained User Authentication on Smart Phones Through Ocular Images Using FHOG and SVM

Recently, perioocular region is gaining popularity amongst the biometric community as an alternative approach to other ocular modalities. This is due to easy imaging and localization from facial images. Therefore, the periocular based user authentication techniques suit well for smart phones. This study aims to combine the image processing and machine learning approaches towards designing a user authentication system for smart phones. For doing so, the study uses viola Jones eye detector for perioocular region localization from the facial images taken with smart phone cameras. The study examines the potential of Felzenszwalb's HOG (FHOG) features combined with support vector machines (SVM) to authenticate the legitimate user of the smart phone. The empirical evaluation of the proposed scheme exhibits promising results on publicly available MICHE – I ocular dataset.