Emotion classification from facial expressions is a significant research area in pattern recognition and artificial intelligence, with wide-ranging applications such as human-computer interaction and behavioral analysis. This study aims to develop a reliable emotion classification system using static images from the FER2013 dataset. A Convolutional Neural Network (CNN) model is implemented as the primary method, initially without preprocessing, followed by the integration of preprocessing techniques to enhance model performance. These techniques include face detection and illumination adjustment, which contribute to generating more representative feature inputs. Feature extraction is performed to optimally identify prominent facial regions such as the jaw, mouth, eyes, nose, and eyebrows. The experimental procedure involves training the CNN model for 33 epochs and evaluating its performance using standard metrics such as accuracy, precision, recall, and F1-score. The results show that the proposed method achieves an average accuracy of 0.9688, a precision of 0.9687, a recall of 0.9688, and an overall Fl-score of 0.9687. Based on these findings, this study recommends the incorporation of preprocessing steps to improve system robustness, particularly for real-time or unconstrained environment applications.
Badri et al. (Fri,) studied this question.