Performance Evaluation of Facial Expression Recognition Using CNN and DRLBP

Abstract

Facial expression recognition is a fundamental problem in computer vision with wide-ranging applications in emotion analysis, human–computer interaction and affective computing. This study proposes an effective facial expression recognition framework that integrates a Convolutional Neural Network (CNN) classifier with the Dynamic Regional Local Binary Pattern (DRLBP) feature extraction technique to enhance recognition accuracy and robustness. The proposed approach follows a structured methodology, beginning with the acquisition of a diverse dataset comprising facial images representing multiple emotional expressions along with their corresponding class labels. Subsequently, discriminative facial features are extracted using the DRLBP algorithm. Unlike conventional Local Binary Pattern (LBP) methods, DRLBP dynamically identifies and analyzes expressive regions of the face based on facial landmarks or key facial points, thereby improving the representation of emotion-related facial variations. The extracted features are then supplied to the CNN model, which efficiently learns hierarchical and high-level representations for classification. The performance of the proposed system is evaluated using standard metrics such as accuracy, precision, recall and F1-score. By effectively combining the powerful feature-learning capability of CNNs with the region-adaptive feature extraction strength of DRLBP, the proposed method achieves reliable and robust facial expression recognition, demonstrating strong potential for real-world applications.