Facial expression recognition using a non-exclusive learning search-Fossa optimization algorithm with a convolutional neural network
Facial expression recognition (FER) is the process of detecting and identifying human emotions based on facial movements and visual cues. It analyzes facial regions, particularly the eyes and mouth, to recognize expressions such as fear, anger, and joy. However, recognizing facial expressions from images remains challenging due to variations in illumination, head orientation, and individual facial characteristics. In this research, a non-exclusive learning search-Fossa optimization algorithm integrated with a convolutional neural network (NELS-FOA-CNN) is proposed to select the most relevant features for accurate FER. In the conventional FOA, NELS is incorporated to enhance the exploration of the solution space, thereby facilitating the identification of optimal solutions and reducing the likelihood of becoming trapped in local optima. A CNN is employed for FER to learn spatial hierarchies of facial features and capture local patterns, such as textures and edges, that are useful for distinguishing among different facial expressions. A baseline graph convolutional network (GCN) is used to compare and validate the performance of the proposed NELS-FOA-CNN. The proposed NELS-FOA-CNN achieves accuracies of 95.80%, 71.23%, and 69.36% on the Real-world Affective Faces Database (RAF-DB), AffectNet-7, and AffectNet-8, respectively, demonstrating improved performance compared with the baseline GCN.