Facial expression recognition using a two stage neural network

An approach based on neural networks for facial expression classification is presented. Two stages of neural networks (NN) are used for expression classification. The first stage neural network based on Hebbian learning is used for estimating principal components (PC) for dimension reduction. The second stage neural network based on supervised competitive learning is used for classifying expressions as normal, happy or surprised. The performance analysis of the proposed algorithm is validated by tabulating confusion matrices and thereby plotting receiver operating characteristics (ROC) curves. Classification rates of 100% and 91.67% are achieved respectively for classification and generalization. The effectiveness of the proposed classification algorithm is highlighted by comparing its performance with another algorithm based on linear discriminant analysis (LDA).