Developing procedures and software for correcting artifacts in motion data

Abstract

Kinesiology relies upon accurate, reliable recordings of movements. Two common techniques for obtaining such recordings are sonomicrometry and optical marker-based motion capture. Although commonly used, sonomicrometry and motion capture are often limited by the presence of artifacts that require correction. Unfortunately, the standard approach to artifact correction in kinesiology is a manual, time-intensive process that requires expertise and painstaking effort to verify. Therefore, to improve the efficiency and consistency of motion data analysis, we have developed procedures and software tools for the correction of artifacts in sonomicrometry and motion capture. Our work on sonomicrometry provides a framework for artifact correction that is more transparent, easier to use, and requires fewer and more limited manipulations from the user compared to existing approaches. Our work on motion capture provides a biomechanically-principled approach to cleaning motion capture data using a dynamic Bayesian network.