Gait impairments are common in Parkinson’s disease, and cueing strategies can help improve mobility. Augmented Reality (AR) headsets offer personalized cues and can capture gait data, but their validity in varied, clinically relevant walking conditions has not been fully tested. This study evaluates the validity and reliability of gait and balance measures from AR headsets by comparing them to a gold-standard motion capture system.
This repository includes a processed dataset in .JSON format containing raw 3D position and orientation data from the AR headsets and the motion capture system.
The motion capture system used is a Vicon system, including 12 Vicon Vero 2.2 camera's, capturing data at 100Hz. A Plug-in Gait lower body model was used, with 4 additional markers on the AR device as seperate segment.
AR headsets used are Magic Leap 2 (OS 1.12.0), capturing positional and rotational data (6-DOF) at ~60Hz.
All participants (n=24) gave informed consent under the study approved by METC-Z (METCZ20250060).
Vicon Nexus, 2.17
Python, 3.12
MATLAB, R2023b
Magic Leap 2 OS, 1.12.0
The dataset can be loaded using the provided example code. JSON files can be read in Python using the built-in json module and in MATLAB using the jsondecode function. Adjust filename based on actual data.
Python:
with open("filename.json", "r") as f:
data = json.load(f)
MATLAB:
data = jsondecode(fileread('filename.json'));
Each JSON file contains a time series of measurements. Each entry includes AR device data with time, x position, y position, z position, pitch, roll, and yaw. Vicon motion capture data includes x, y, and z positions for each marker.
Explanation of tests and data are included in the README.md.
