Facial Expression Recognition Based on Improved Non-negative Matrix Factorization
Tian Jian-hua · Computer Knowledge and Technology · 2011
This paper presented a facial expression recognition method based on improved non-negative matrix factorization with sub-region occlusion judgment.Given a still image containing facial expression information,expression feature vectors are extracted by non-negative matrix factorization firstly.Secondly,facial expressions are classified by using a maximum correlation classifier.Experiment on Cohn-Kanade facial expression database shows that the method improves the expression recognition for both non-occluded and partially occluded faces.