Face recognition: dealing with expression variations

Supriya D. Kakade and Kanade S. G

Face Recognition is integral part of biometrics. It has been a fast rising, challenging and attention-grabbing area in real time applications i.e. image analysis, pattern recognition, etc. Facial expressions is one of the most critical sources of variations in face recognition, especially in the common case where for enrollment only a single sample per person is existing. For a consistent authentication system, methods that improve the accurateness in the occurrence of such variations are still required. In this paper, we deal with this problem with an analysis by-synthesis-based method in which a number of synthetic face images with different expressions are formed. For this purpose, 3D mask is generated for each user based on respective landmark points. The role of these extra images in terms of the recognition performance is evaluated with different techniques on face recognition magnificent challenge and Bosphorus 3D face databases. The project is implemented using more than one sample of expression for a single person to get more accuracy in face recognition

Download PDF: