Noise Reduction in Human Motion-Captured Signals for Computer Animation based on B-Spline Filtering

Area of Science:

• Computer Graphics
• Human-Computer Interaction
• Signal Processing

Background:

• Motion capture records natural human movements for animation and virtual reality (VR).
• Raw motion data often contains noise from the capturing process, necessitating signal processing.
• Effective signal smoothing is crucial for applying captured motion to digital characters.

Purpose of the Study:

• To analyze common signal smoothing methods for motion capture data.
• To compare the performance of different smoothing techniques based on defined metrics.
• To identify an optimal filtering method for processing noisy motion signals.

Main Methods:

• Evaluation of several established signal smoothing algorithms.
• Application of B-spline-based least squares filtering.
• Comparison of smoothed signals using quantitative smoothness metrics.

Main Results:

• Identified B-spline-based least squares filtering as a superior method.
• Demonstrated high-quality output generation with this technique.
• Showcased minimal parameter adjustment requirements for diverse motion signals.

Conclusions:

• B-spline-based least squares filtering provides effective noise reduction for motion capture data.
• This method achieves predetermined continuity and high-quality motion signal processing.
• It offers a robust solution for preparing motion data for animation and VR applications.