Optimizing dance motion reconstruction using a two-dimensional matrix approach with hybrid genetic and fuzzy logic...

Area of Science:

• Biomechanics and Motion Capture
• Computational Intelligence and Machine Learning
• Rehabilitation Engineering

Background:

• Motion capture for dance movement reconstruction faces challenges with body morphology, clothing, and nonlinear human dynamics.
• Existing methods struggle with intricate motions, exhibiting parameter sensitivity and local optima issues.

Purpose of the Study:

• To develop a precise method for reconstructing complex human dance movements.
• To address limitations in existing motion capture techniques for nonlinear biomechanical patterns and missing data.

Main Methods:

• Development of the Two-Dimensional Matrix-Calculation (TDMC) model integrated with a Hybrid Genetic Algorithm and Fuzzy Logic Differential Evolution (HGA-FLDE).
• Leveraging Riemannian geometry and adaptive optimization for biomechanical nonlinear motion patterns.
• Comparative analysis with Long Short-Term Memory (LSTM), Support Vector Regression (SVR), Kinect Sensors (KS), and Evolved Deep Gated Recurrent Unit (EDGRU) models.

Main Results:

• The TDMC-HGA-FLDE model achieved a high accuracy of 0.95 at 60 nodes, surpassing baseline models.
• It demonstrated a minimum error rate of 0.39 at 20 nodes.
• In a dance therapy use case for lower limb rehabilitation with incomplete IMU data, TDMC-HGA-FLDE achieved 0.94 accuracy and 0.22 MSE.

Conclusions:

• The hybrid TDMC-HGA-FLDE approach significantly enhances the precision and realism of dance motion reconstruction.
• This method offers substantial improvements for motion capture and rehabilitation applications, particularly in dance therapy.
• The model effectively handles nonlinear dynamics and missing data in biomechanical motion capture.