Authentication Based on Periocular Biometrics and Skin Tone

Face images with masks have a major effect on the identification and authentication of people with masks covering key facial features such as noses and mouths. In this paper, we propose to use periocular region and skin tone for authenticating users with masked faces. We first extract the periocular region of faces with masks, then detect the skin tone for each face. We then train models using machine learning algorithms Random Forest, XGBoost, and Decision Trees using skin tone information and perform classification on two datasets. Experiment results show these models had good performance.