Facial expression recognition using general regression neural network

Automated facial expression recognition using machine learning techniques is a preliminary step for the future work on the automated emotion detection systems to be used in the real world. In this paper, facial expression recognition using Efficient Local Binary Pattern (LBP) for feature extraction and artificial neural network (ANN) for classification is presented. Facial feature vectors are obtained using Efficient LBP algorithm by considering the blocks of four varied sizes (256 × 256, 128 × 128, 64 × 64 and 32 × 32) of the image dataset. The multiclass classification of the image dataset into the six basic universal emotions is carried out using ANN. We have implemented General Regression Neural Network (GRNN) for classification in this experiment. GRNN trains the network faster and does not require iterative training procedure. The free parameter of the network, spread constant is optimized to reduce the mean square error and thus the efficiency of recognition is improved as seen from the experimental results. The proposed algorithm is tested using widely used standard database such as Japanese Female Facial Expression Database, Taiwanese Facial Expression Database, Cohn-Kanade Expression Database and Indian face database of students. The proposed algorithm with the optimum window sizes of 64×64 improves the recognition rate.