Facial Expression Recognition Based on Constrained Local Models and Support Vector Machines

This paper presents a face expression recognition algorithm using Constrained Local Model (CLM). CLM is facial alignment method that is based on Active Shape Models (ASM) and Active Appearance Models (AAM). It takes the advantages of both of them and gains high accuracy. To distinguish different expression states, we use CLM model parameters that describe shape deformation in a compact form. These parameters form feature vectors for training Kernel Support Vector Machine (KSVM) classifier. The experimental results over Cohn-Kanade Extended Facial Expression (CK+) database show improvement of the recognition rate in comparison to some existing methods, suggested by other authors.

• Nikolay Neshov
• Ivo Draganov
• Agata Manolova

International Scientific Conference on Information, Communication and Energy Systems and Technologies (ICEST), 2015.

http://eprints.uklo.edu.mk/4236/