Advancements in computational emotion recognition: a synergistic approach with the emotion facial recognition dataset and RBF-GRU model architecture

Subhranil Das, Rashmi Kumari, Raghwendra Kishore Singh

Abstract: In this study, Radial Basis Function-Gated Recurrent Unit (RBF-GRU) has been proposed for classifying five different facial expressions, such as, Anger, Fear, Sadness, Happiness, and Neutrality. Also, we have developed Emotional Facial Recognition (EPR) dataset where that dataset contains 30,000 different images by taking consideration of Indian scenarios. For this developed model, it has been well designed for robust training and evaluating the EPR dataset. The advantage of using this model is to deliver its capabilities if any minor perturbation are present. For evaluating the performance of the model, confusion metrics parameters have been utilized which are Classification Accuracy, Precision, F1-Score, Sensitivity, and Specificity for classifying five different emotional faces. Moreover, the efficacy of the proposed RBF-GRU model has been evaluated in the existing datasets, such as CK + , FER-2013, and JAFFE. In addition to that, the effectiveness of the proposed model RBF-GRU has been compared with four existing Deep Learning algorithms where the model has achieved all the parameters at a higher note which confirms the reliability in the different emotions tasks.

External IDs:dblp:journals/saem/DasKS25