MARKERLESS MOTION TRACKING

Markerless motion caption with a single camera is a promising approach for skeleton tracking. Nowadays, used within video gaming it is a possible system for clinical gait analysis. This project aims to assess the reliability and validity of an Intel® RealSenseTM camera compared to a reference system for gait analysis in running sports. Inertial Measurement Units were used as the reference standard and a Python script based on the Nuitrack SDK was implemented to process the optical data.
As part of a feasibility study, seven healthy participants were tracked with both systems during walking and jogging indoors and outdoors. The knee, hip, shoulder, and elbow angles were tracked simultaneously and processed to be statistically analyzed.
Although the optical system was able to track skeletons in outdoor and indoor settings, the results of the outdoor settings were too poor to be analyzed. The indoor measurement results found a moderate ability of the optical system to approximately match the knee, shoulder, and elbow angles measured with the reference system. The hip angles generated the highest errors and the lowest correlation. In conclusion, the Intel® RealSenseTM camera and Nuitrack SDK are not yet reliable and valid tools for gait analysis in clinical settings and need improvement hardware- and software-wise.

Used Hardware

Used Software

Implementation using Python

The following angles were calculated by establishing several vectors taking the tracked coordinates and using the dot product: shoulder angle α, right hip H, right shoulder S and right elbow E. All angles were then stored as CSV files. The recorded videos, RGB and depth, were displayed and stored as single frames as PNG.

Experimental Setup

Indoor setting:

One laptop equipped with the EMG Motion Tools Software connected to the MYON Aktos Reciever.