Reflectance spectra based skin and non-skin classification

Face recognition systems have been widely studied and integrated in various applications. Under certain spoofing attacks such as displaying high-resolution photos/videos or wearing masks made up of non-skin materials, they fail. In this work, we propose a new approach of skin and non-skin classification, based on spectral reflectance, providing more discriminative signatures. This approach allows synthesis of biological skin features based on photo-biophysics model. Here, a simple 2-layered skin model is applied which is proofed to be effective in skin and non-skin classification both from theoretical explanations and experimental results. This approach is compared to direct spectral reflectance analysis and classification. Experimental results can reach a classification rate up to 96.4%.