Neural Network Classifier for Human Emotion Recognition from Facial Expressions Using Discrete Cosine Transform

This research aims at developing "humanoid robots" that can carry out intellectual conversation with human beings. The first step in this direction is to recognize human emotions by a computer using neural network. In this paper all six universally recognized basic emotions namely angry, disgust, fear, happy, sad and surprise along with neutral one are recognized. Multilayer perceptron (MLP) and generalized feed forward neural network (GFFNN) are employed and their performance is compared. Discrete cosine transform (DCT) and statistical parameters are used for feature extraction. The authors achieved 100% recognition rate on training data set (seen examples) and test data set (unseen examples).