Enhanced AlexNet with Gabor and Local Binary Pattern Features for Improved Facial Emotion Recognition.

Area of Science:

• Computer Science
• Artificial Intelligence
• Machine Learning

Background:

• Facial emotion recognition (FER) is crucial for human-machine interaction and AI systems.
• Real-world applications often face hardware limitations, necessitating efficient FER models.
• Integrating cognitive and emotional intelligence enhances machine engagement with humans.

Purpose of the Study:

• To propose an enhanced deep learning model for facial emotion recognition (FER).
• To address the challenges of low hardware specifications in real-world FER applications.
• To improve the accuracy and adaptability of FER systems.

Main Methods:

• Leveraged deep learning advancements for FER model development.
• Utilized Gabor and Local Binary Pattern (LBP) for texture feature extraction.
• Integrated features into a modified AlexNet architecture.

Main Results:

• Achieved 98.10% accuracy on the FER2013 dataset and 93.34% on the RAF-DB dataset.
• Demonstrated high precision, recall, and F1-scores on both datasets.
• Showcased model robustness and performance under various operational conditions.

Conclusions:

• The proposed FER model offers high-precision emotion recognition.
• The model is suitable for deployment in resource-constrained environments.
• This research contributes to more effective human-machine interactions through advanced AI.