Facial micro-expression recognition based on motion magnification network and graph attention mechanism

Area of Science:

• Computer Science
• Artificial Intelligence
• Biomedical Engineering

Background:

• Micro-expressions reveal genuine emotions but are challenging to recognize due to subtle facial muscle movements.
• Existing methods struggle to accurately detect these fleeting emotional indicators.

Purpose of the Study:

• To develop an advanced network for accurate micro-expression recognition.
• To enhance the detection of subtle facial muscle motions for improved emotion analysis.

Main Methods:

• Proposed a Swin Transformer-based network (ST-MEMM) for micro-expression motion magnification, amplifying subtle movements.
• Introduced a Graph Attention Mechanism-based network (GAM-MER) for recognition, optimizing facial landmarks and prioritizing key features.
• Utilized CASME II and SAMM datasets for experimental validation.

Main Results:

• The proposed Graph Attention Mechanism-based Motion Magnification Guided Micro-Expression Recognition Network (GAM-MM-MER) demonstrated high accuracy.
• Achieved significant superiority over existing state-of-the-art methods in micro-expression recognition.
• Ablation studies confirmed the robustness and efficacy of the proposed network.

Conclusions:

• The GAM-MM-MER network effectively enhances subtle muscle motions and focuses on critical facial landmarks.
• The proposed method offers a significant advancement in the field of micro-expression recognition.
• The network proves robust and effective for accurately identifying genuine emotions from micro-expressions.