Adaptive Motion Data Representation with Repeated Motion Analysis.

Area of Science:

• Computer Graphics and Animation
• Digital Signal Processing
• Human-Computer Interaction

Background:

• Motion capture data is crucial for realistic animation and virtual environments.
• Existing methods often struggle with large data sizes and efficient compression.
• Human motion exhibits inherent characteristics like repetition and spatiotemporal coherence.

Purpose of the Study:

• To develop an efficient representation method for motion capture data.
• To exploit near-repeated characteristics and spatiotemporal coherence in human motion.
• To achieve substantial compression gains with high-quality reconstruction.

Main Methods:

• Extracting similar motion clips of variable lengths and speeds.
• Employing repeated motion analysis to identify referred and repeated clip pairs.
• Approximating subspace-projected motions using interpolated functions with adaptive quantization.

Main Results:

• The proposed feature-aware method demonstrates high computational efficiency.
• Substantial compression gains were achieved.
• Comparable reconstruction and perceptual errors were maintained.

Conclusions:

• The novel representation method effectively utilizes human motion characteristics for compression.
• The approach offers a balance between compression efficiency and data fidelity.
• This method provides a computationally efficient solution for motion capture data management.