Cleaning and labeling gait data: an automatic marker-wise algorithm

Abstract

Optic motion capture (MoCap) is a process that captures the location of reflective markers placed on the patient's skin to analyze their biomechanics through different models. The record of these markers can present occlusions and noise that makes the raw data cleaning and its labeling impossible to be automatized. This process is time consuming for the operator. There is no solution in the literature that addresses both problems simultaneously. In this work we propose an algorithm that cleans up ghost and spurious markers, fills in their gaps and labels them based on a previously loaded template. Its robustness with simulated data was studied by increasing the noise of the acquisitions and the gaps with lack of information. The results obtained are promising, demonstrating the robustness of the algorithm and its feasible application in a clinical environment, reducing processing times.