Fusion of motion smoothing algorithm and motion segmentation algorithm for human animation generation.

Area of Science:

• Computer Graphics
• Human Animation Generation
• Virtual Reality

Background:

• Current human animation technologies struggle with large datasets and capturing subtle motion transitions, leading to unnatural and unrealistic animations.
• Existing methods often lack fluency and realism due to difficulties in processing motion transitions and dynamic changes.

Purpose of the Study:

• To enhance the naturalness and diversity of human animation generation.
• To overcome limitations in existing animation technologies regarding data dependence and motion transition processing.

Main Methods:

• Employed a tree-level model with a bidirectional unbiased Kalman filter for noise reduction in motion data, improving motion capture accuracy.
• Utilized discriminant analysis based on sparse reconstruction and multi-scale temporal association segmentation to identify key motion segments adaptively.
• Combined motion smoothing and motion segmentation algorithms for improved animation generation.

Main Results:

• Achieved high accuracy in motion segmentation: 0.96 for coarse-grained and 0.91 for fine-grained.
• Demonstrated superior performance in color fidelity, detail representation, motion fluency, and action authenticity compared to prior art.
• Reported average user satisfaction exceeding 0.85, indicating enhanced character expressiveness and realism.

Conclusions:

• The proposed method significantly improves the naturalness, diversity, and realism of human animation.
• This research offers advancements for computer graphics, virtual reality, and augmented reality applications.
• The technique effectively addresses challenges in motion capture accuracy and dynamic change processing for animation.