Multiple Attention Mechanism Graph Convolution HAR Model Based on Coordination Theory

Area of Science:

• Computer Vision
• Artificial Intelligence
• Human-Computer Interaction

Background:

• Human action recognition (HAR) is crucial for understanding human behavior in various applications.
• Limb coordination in human kinematics provides rich information for motion analysis.
• Existing HAR algorithms often require multifaceted attention to different joints.

Purpose of the Study:

• To develop an improved HAR algorithm focusing on limb coordination.
• To enhance the model's attention to critical joints during motion analysis.
• To increase the accuracy of human action recognition.

Main Methods:

• Proposed a novel HAR algorithm incorporating two synergistic attention modules.
• These modules are designed to extract coordination characteristics from motion data.
• The algorithm dynamically adjusts attention to important joints.

Main Results:

• Experimental validation on public datasets (NTU-RGB+D, Kinetics-Skeleton) demonstrated significant improvements.
• The proposed dual-attention mechanism enhanced recognition accuracy.
• The model effectively captured complex limb coordination patterns.

Conclusions:

• The novel dual-attention HAR algorithm effectively captures limb coordination.
• This approach leads to superior recognition accuracy compared to baseline methods.
• The findings have implications for advanced human behavior analysis systems.