Soft Biometrics Authentication: A Cluster-Based Skin Color Classification System

Introduction

Adapted multi-biometrics consists of introducing ancillary data into the biometric authentication process to improve recognition performance. These ancillary data can come from the user, the sensor or the acquisition environment. If the ancillary data come from the user, this is called soft biometrics. Djara et al. (2019) have developed a typology of biometric adaptation. Within the wide range of biometric adaptation data, soft biometric data are increasingly being researched. Soft biometrics includes, but are not limited to, height, weight, gender, skin color, hair color, eye color, etc. (Dantcheva, 2011). Skin color is one of the most remarkable means of human differentiation and has been widely used to define human races (Jablonski & Chaplin, 2000). Among these soft biometric modalities, skin color appears to be relevant in terms of discriminatory power. As skin color is a soft biometric modality, it cannot by itself reliably authenticate an individual, but it can be merged with other pure modalities to create a robust multi-modal system. We propose through this manuscript a new approach of authentication by skin color classification.