Facial Emotional Recognition Using Faster Regional Convolutional Neural Network with VGG16 Feature Extraction Model

Facial Emotion Recognition (FER) poses a significant challenge when training deep learning (DL) models like Artificial Neural Network (ANN) and machine learning (ML). The primary issue in FER is the detection of the face, the creation of bounding boxes, and the classification of emotional expressions. This paper proposes the use of Faster Regional Convolutional Neural Network (RCNN) for facial recognition and employs a median filter to filter the image dataset. The model leverages the Facial Emotional Recognition (FER 2013) dataset for training and predicting accurate classifications. The Median Filter is used to eliminate noise in the image, such as Salt and Pepper noise, and data augmentation is utilized to generate synthetic data for addressing minor imbalances in classes. Subsequently, the VGG 16 architecture is implemented for precise feature extraction to reduce complexity in the classification process. This paper, achieved accuracy, precision, recall, specificity, and f1-score values of approximately 78.22%, 75.40%, 80.20%, 85.90%, and 71.40%, respectively, when compared to other classification methods like Convolutional Neural Network, ResNet 50, Divide-And-Conquer-Based learning, and Resnet 18.