Biometric Face Recognition Based on Landmark Dynamics
Andrea Francesco Abate, Lucia Cimmino, Fabio Narducci, Chiara Pero · 2020
Face identification is one of the most widely adopted approach in several real scenarios. Unfortunately, in presence of occlusions or non-cooperative subjects nontrivial issues arise. This work explores a complementary approach to biometric face recognition, that is the dynamics of facial landmark during phonation while pronouncing sentences. These dynamics (both periocular and labial area) can be seen as a kind of signature that uniquely and unmistakably identifies an individual. The experimental results focus the attention on periocular area, which represents one of the most discriminating physiological characteristics after the whole face. Preliminary experimental results conducted on a public dataset and considering 14 periocular features showed an accuracy slightly below 80%, which confirms the robustness in uncontrolled scenarios.