An efficient method of key-frame extraction based on a cluster algorithm.

Area of Science:

• Computer Science
• Robotics
• Animation

Background:

• Motion capture data is crucial for animation and robotics.
• Efficiently extracting key information from motion capture sequences is challenging.
• Existing methods often require manual parameter tuning.

Purpose of the Study:

• To propose a novel, unsupervised key-frame extraction method for motion capture data.
• To develop an adaptive clustering algorithm that eliminates the need for user-specified parameters.
• To reliably and efficiently summarize complex motion capture sequences.

Main Methods:

• The method employs an unsupervised clustering algorithm based on adjacent frame similarity.
• Initial clustering adaptively determines thresholds for subsequent analysis.
• The ISODATA dynamic clustering algorithm identifies frames closest to class centers as key-frames.

Main Results:

• The proposed method successfully extracts key-frames from diverse motion types without parameter input.
• Evaluations show reliable and efficient summarization of motion capture data.
• Comparisons with previous methods demonstrate competitive or superior performance in motion reconstruction and error metrics.

Conclusions:

• The novel unsupervised clustering approach offers an effective solution for key-frame extraction in motion capture.
• The adaptive and parameter-free nature of the algorithm enhances its usability and applicability.
• This method provides a robust tool for managing and analyzing large motion capture datasets.