Goniometry-based Glitch-Correction Algorithm for Optical Motion Capture Data

Nowadays , motion capture technology is used in productions of all levels in 3D Animation. The concept on which this technology is based on, consists of the elaboration of 3D models from numerical data taken by a set of sensors (for example infrared cameras) interpreted by a software. The problem with this technology is that the processing of data generated by these sensors is not always accurate, causing loss of positional data which results in errors called glitches, that produce corrupted 3D models. In this work, a goniometry-based algorithm to detect, locate and correct the glitches generated from optical motion capture data is presented. Based on the classification of angular measures of the articular physiology in humans, a pattern recognition approach was used to construct the algorithm. The proposed algorithm produces average F1-scores of 0.956 using synthetical data, and produces natural results in most of the cases for real data.