Facial emotion plays a vital role for human interactive communication and also used in numerous real applications. Facial expression identification from frontal still images has in recent times become a hopeful investigation area. Their applications include human-computer interface, human emotion examination robot control, driver state surveillance and medical fields. This paper aims to develop emotion classification scheme to identify seven dissimilar facial emotions, such as surprise, sad, neutral, happy, fear, disgust and anger by using JAFFE database. Two different approaches of feature selection and extraction have been used for generation of optimal feature vector. LBP and 2D-DCT coefficients are employed in addition to image statistics, texture and entropy parameters. In order to reduce the high dimensionality of the inputs, the principal component analysis has been used and significant reduction in the input-dimensionality has been achieved. The single hidden layer feed forward neural network has been used as a classifier in order to classify different emotions from frontal facial images. Three learning algorithms such as resilient backpropagation, scaled conjugate gradient and gradient descent algorithm with momentum and adaptive learning rate have been compared. It has been observed that our meticulously designed LBP based hybrid feature vector and a single hidden layer neural network containing only 70 neurons in the hidden layer trained with gradient descent algorithm with momentum and adaptive learning rate delivers the maximum average overall classification accuracy of 97.2%, which has not been reported so far in the literature for the said database. The proposed neural network is very compact, as it is comprised of only 5,607 connection weights including biases.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
customersupport@researchsolutions.com
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.