DeXpression: Deep Convolutional Neural Network for Expression\n Recognition

Abstract

We propose a convolutional neural network (CNN) architecture for facial\nexpression recognition. The proposed architecture is independent of any\nhand-crafted feature extraction and performs better than the earlier proposed\nconvolutional neural network based approaches. We visualize the automatically\nextracted features which have been learned by the network in order to provide a\nbetter understanding. The standard datasets, i.e. Extended Cohn-Kanade (CKP)\nand MMI Facial Expression Databse are used for the quantitative evaluation. On\nthe CKP set the current state of the art approach, using CNNs, achieves an\naccuracy of 99.2%. For the MMI dataset, currently the best accuracy for emotion\nrecognition is 93.33%. The proposed architecture achieves 99.6% for CKP and\n98.63% for MMI, therefore performing better than the state of the art using\nCNNs. Automatic facial expression recognition has a broad spectrum of\napplications such as human-computer interaction and safety systems. This is due\nto the fact that non-verbal cues are important forms of communication and play\na pivotal role in interpersonal communication. The performance of the proposed\narchitecture endorses the efficacy and reliable usage of the proposed work for\nreal world applications.\n