Face analysis and demographic attributes identification play an important role in many applications and have gained increasing attention. Appearance-based methods using passively acquired images can yield satisfying results on gender classification, but there is still room for improvements in ethnicity and age classifications. Alternatively, some researchers focused on actively acquired hyper-spectral images, and use physics-based skin models to synthesise histological parameters for ethnicity classification, but these are little studied. In this paper, we first introduce several state-of-the-art physics-based skin models including implementation, their advantages and drawbacks; then we present their applications in gender, skin type and age classifications.