Authors : Priyanka A. Gavade, Vandana S. Bhat and Jagadeesh Pujari
Abstract: Facial expression is the nonverbal way to express the human intensions and emotions. Facial Expression Recognition (FER) intends to understand and analyze the facial behavior of humans such that it has become an active research area in the field of pattern recognition, artificial intelligence and computer vision. Various FER methods are developed for classifying the facial expression in video sequences but to extract the discriminative video features from the facial expression images results a key challenging issue in FER system. Hence, an effective FER method is designed using proposed Taylor-Chicken Swarm Optimization-based Deep Generative Adversarial Network (Taylor-CSO based Deep GAN) for the recognition of facial emotions. However, the proposed method named Taylor-CSO is derived by the integration of Taylor series with Chicken Swarm Optimization (CSO), respectively. The process of Illuminant Invariant Local Binary Pattern (IILBP) is made by employing the LBP descriptor to the facial object. Based on the feature matrix, the process of FER is accomplished using Deep GAN. However, the proposed approach achieved the accuracy, precision and recall of 0.8846, 0.8996 and 0.8952 with respect to training data.
Priyanka A. Gavade, Vandana S. Bhat and Jagadeesh Pujari, 2021. Taylor-Swarm Ganet: Learning Illumination Invariant Feature Descriptor For Facial Expression Recognition using Deep Generative Adversarial Network. Journal of Engineering and Applied Sciences, 16: 208-221.