Improved Motion Capture Processing for High-Fidelity Human Models Using Optimization-Based Prediction of Posture and Anthropometry

Abstract

Improving motion capture processing onto a high-fidelity digital human model is an important research area. Although there has been significant research in this field, little work has been done to determine posture and anthropometry simultaneously with the intent of visualizing the data on high-fidelity human models. Many existing techniques are less accurate when applying processed data to a digital model for biomechanical analysis. This paper presents a novel approach that estimates posture and anthropometry using optimization-based posture prediction to determine joint angles and link-lengths of a digital human. By including anthropometric design variables, this approach introduces flexible handling of innate variance in subject-model measurements without need for pre- or post-processing. This produces a more realistic motion and exhibits anthropometric measurements closer to those of the original subject, resulting in a new level of biomechanical accuracy that allows for analysis of a processed motion with a higher degree of confidence.