A Rigid Structure Matching-Based Noise Data Processing Approach for Human Motion Capture

A noise data processing algorithm based on rigid structure matching for human motion capture (MoCap) is presented in this paper. The noise model is related to parts of human body, but not to trajectories of markers as most used in traditional approaches. Three issues are addressed in this method: rigid structure matching, semantic restriction and noise reduction, and they are three steps of data processing. First, human body is divided into seven main parts: head, chest, waist, hands and legs, and all parts´ candidate parts are generated by algorithm of rigid structure matching from motion data. Missing markers can be automatically reconstructed by existing markers on a same rigid structure. Subsequently, one group of parts with most possibility is determined as initial processed data by semantic restriction. Finally, an impulse noise model is found based on the analysis of data, and the noise has been filtered. At the end, the main conclusions of this research are drawn, and the advantages and limitations of this approach are discussed respectively. The algorithm presented can be easily implemented and performed, and the processed motion data can be expediently used in human body animation. The efficiency of the method is illustrated by the experiments.