Human Facial Expression Recognition with Convolution Neural Networks

Facial expression recognition (FER) is an active area in machine learning research, where human-machine interaction is prevalent for developing applications such as health care, gaming, and augmented reality. Many attempts have been made to find efficient solutions capable of improving the recognition accuracy. In this paper, we study how machine learning methods, such as convolution neural networks (CNNs), can improve the FER accuracy in biometric applications. We describe our approach and show that the proposed solution can improve the accuracy on the FER2013 data which include real facial images assigned to the seven facial expressions categories.