Face Recognition using Belief Propagation Algorithm for Varying Expression

S. Okuno, T.T. Son, and S. Mita (Japan)


Belief propagation algorithm (BPA), Active shape model(ASM), facial expression recognition, face recognition.


The purpose of this work is to establish a robust face recognition system that is capable of handling changes of face status in human-like attitude. The proposed face recognition system first estimates the expression for one input of face images and provides the classification result into a proper database for future matching. Face identification is then performed for each facial expression. The belief propagation algorithm is applied to both expression estimation and recognition. Our proposed system was tested with a CMU PIE database that included various expressions. The system achieved a recognition rate of 95.0% when we used only 1 image per person for each expression in our training data.

Important Links:

Go Back