Development of Facial Expression Classifier using Neural Networks

A person's emotional and mental well being, together with the age, sex, race, can be easily depicted by one's face. A crucial role is played by facial expressions in day-today social interactions. An individual's emotional level as well as behavioral manners can be interpreted by these expressions. Facial expression classifier is a evolving, demanding and curious problem in computer vision. It has its potential applications in the field of robotics, behavioral science, human computer interaction, video games etc.. It assists in building more intelligent systems which have better ability to interpret human emotions. In this paper, a facial expression classifier is proposed based on Convolution Neural Networks (CNN). CNNs are biologically-inspired variants of multi-layer preceptor (MLP) networks. They use an architecture which is particularly well suitable to classify images. Detection of facial expression can be enhanced by utilizing the ability of learning facial features. The results showed that deep CNNs are apt for this task. We have used CNN to interpret and classify the facial expression label into seven different classes: anger, happiness, fear, sadness, disgust, surprise and neutral.