Facial expression recognition based on deep learning

Area of Science:

• Computer Vision
• Artificial Intelligence
• Machine Learning

Background:

• Facial expression recognition is crucial for human-computer interaction in autonomous driving and robotics.
• Deep learning, particularly convolutional neural networks, dominates computer vision tasks.

Purpose of the Study:

• Propose an occluded expression recognition model using a generative adversarial network.
• Enhance facial expression recognition in real-world, challenging environments.

Main Methods:

• Summarize deep learning methods for facial expression recognition over the past decade.
• Categorize and analyze static and dynamic facial expression recognition techniques.
• Compare algorithm performance on common expression databases.

Main Results:

• Deep neural networks learn discriminative features for automatic facial expression recognition.
• Current systems struggle with overfitting due to limited data and environmental interferences.
• The developed model aims to overcome these limitations.

Conclusions:

• Integrating multimodal information (audio, 3D depth, physiological data) enhances expression recognition.
• Combining facial action unit and dimension models improves practicality.
• Further research can lead to more robust and applicable facial expression recognition systems.