90 / 2025-05-14 22:00:35

Facial Expression Recognition Using Deep Learning: A Hybrid Approach with ResNet121 and LightGBM

Facial Expression Recognition, Deep Learning, ResNet121, LightGBM, Feature Extraction, Emotion Recognition

Facial expression recognition (FER) plays a vital role in human-computer interaction, social robotics, and mental health monitoring. This paper proposes a novel deep learning framework for FER that combines ResNet121 for deep feature extraction and LightGBM for efficient classification. Traditional FER methods often suffer from low accuracy and poor generalization due to manual feature engineering and limited learning capabilities. To address these challenges, we introduce a hybrid architecture that leverages the power of deep convolutional networks and gradient boosting. Our method first employs ResNet121 to extract high-level discriminative features from facial images, followed by a LightGBM classifier optimized with weight normalization to enhance classification performance. Experiments conducted on the FER13 dataset demonstrate that our approach achieves superior accuracy (99.19%), recall (98%), F1 score (98%), and precision (98%) compared to existing methods. The results highlight the effectiveness of our framework in capturing complex emotional expressions and its potential for real-time applications. Future work will explore the integration of multimodal data and advanced architectures like Transformers to further improve robustness and generalization.