A Dynamic Face Recognition System

Nowadays most face recognition systems arc designed to cope with static environment. In many scenes such as smart surveillance people are in dynamic environment. Exist static recognition system can't work well in dynamic case. This paper propose a probabilistic model of dynamic face recognition. Depending on the probabilistic model we made two improvements to the previous method. Firstly, this paper introduce a parallel framework. Secondly, an effective face recognition algorithm which is based on Deep Learning and Mahalanobis distance metric learning is proposed. Additionally, this paper propose a different method to evaluate and select face recognition algorithm in dynamic case. Theoretical analysis and real world experiments result shows that our system framework can solve the situation of real-time face recognition in dynamic environment.