Sparse Representation Feature for Facial Expression Recognition

Facial expression recognition is a challenging task, because it is difficult to recognize facial expressions of different persons if they are of diverse races and ages. Extracting distinctive feature from original facial image is a critical step for successful facial expression recognition. This paper proposes sparse representation feature for facial expression recognition. First of all, a dictionary is established using training images. Then sparse representation feature is extracted by sparse representation orthogonal matching pursuit method. Finally the extracted features of different expressions are classified by two-hidden-layer extreme learning machine. Facial expression images of both Cohn-Kanade and JAFFE databases are classified using sparse representation feature. Experimental results show that the sparse representation feature is suitable for facial expression recognition.