Leveraging Two Kinect Sensors for Accurate Full-Body Motion Capture

Area of Science:

• Computer Vision
• Human-Computer Interaction
• Biomechanical Analysis

Background:

• Accurate human motion capture is crucial for sports, medicine, and virtual reality.
• Existing monocular depth camera systems struggle with occlusions, limiting pose estimation accuracy.
• Occlusion is a significant challenge in real-world motion analysis.

Purpose of the Study:

• To develop a robust human pose estimation system that overcomes occlusion limitations.
• To enhance the accuracy of motion capture through multi-sensor fusion and temporal analysis.
• To provide a reliable solution for precise human body parameter measurement.

Main Methods:

• Utilized two Kinect sensors for comprehensive human movement data acquisition.
• Implemented a learning analysis approach to leverage temporal information from posture variations.
• Integrated temporal domain constraints to refine pose parameter estimation.
• Developed a complete system for accurate measurement of human body pose parameters.

Main Results:

• Achieved accurate human pose estimation even with significant occlusion.
• Demonstrated the system's capability to handle rapid human movements effectively.
• Experimental results validated the efficacy of temporal information in pose estimation.
• The proposed system shows superior performance compared to monocular systems under occlusion.

Conclusions:

• The dual-Kinect system with temporal analysis provides accurate human pose estimation.
• The approach effectively mitigates challenges posed by occlusion in motion capture.
• This method offers a reliable solution for precise human motion analysis in various applications.
• Future work could explore real-time implementation and broader application domains.