Gender, Skin Type and Age Classifications using Skin Reflectance-based Descriptor

International audienceDemographics attributes such as gender, ethnicity and age play an important role in many applications such as demographic statistics, targeted advertising, medical diagnosis etc. Identification of these attributes has gained increasing attention and has been widely investigated. The state of the art methods can be divided into 2 categories: geometric-based and appearance-based methods. The first method calculate the distances between manually maintained facial landmarks, some useful information may be thrown away; while the second extracts texture or shape information from passively acquired facial images, such methods can yield satisfying results in gender identification, however, they are inefficient in ethnicity and age identifications, especially for intermediate classes. Alternatively, some researchers focused on actively acquired hyper-spectral images, they synthesized histological parameters to classify ethnicity, but few work has been investigated. The aim of this work is to introduce a classification scheme for gender, skin type, age simultaneously for each facial region with a fusion technique, and to analyze the relations with skin thickness