Collaborative discriminative multi-metric learning for facial expression recognition in video

Facial expression recognition in video has been an important and relatively new topic in human face analysis and attracted growing interests in recent years. Unlike conventional image-based facial expression recognition methods which recognize facial expression category from still images, facial expression recognition in video is more challenging because there are usually larger intra-class variations among facial frames within a video. This paper presents a collaborative discriminative multi-metric learning (CDMML) for facial expression recognition in video. We first compute multiple feature descriptors for each face video to describe facial appearance and motion information from different aspects. Then, we learn multiple distance metrics with these extracted multiple features collaboratively to exploit complementary and discriminative information for recognition. Experimental results on the Acted Facial Expression in Wild (AFEW) 4.0 and the extended Cohn-Kanada (CK+) datasets are presented to demonstrate the effectiveness of our proposed method.