Datasets and pre-processing pipelines accompanying the study: Predicting gait kinetics using 3-degrees of freedom acceleration data and artificial neural networks

Published: 6 September 2026| Version 1 | DOI: 10.17632/x7srp5snfk.1
Contributors:
,
, Alexander Hobein,
,
,

Description

This study evaluates whether 3D human gait kinetics can be accurately predicted outside a laboratory setting. The core hypothesis is that wearable linear acceleration data (3 DoF), combined with artificial neural networks (LSTM and MLP), can successfully estimate clinically relevant parameters without the need for resource-intensive camera systems and force plates. The repository contains anonymized time-series datasets from 32 healthy subjects alongside the Python pre-processing pipelines. The 'ReadMe.txt' file explains the scripts and the overall data structure.

Files

Steps to reproduce

Detailed information on how the data was arrived at—including the experimental protocols, instruments, software workflows, and data-gathering methods—is comprehensively explained in the accompanying peer-reviewed publication in Clinical Biomechanics (see the related article link).

Institutions

Categories

Medicine, Artificial Intelligence, Biomedical Engineering, Orthopedics, Biomechanics, Machine Learning, Gait Analysis

Funders

Related Links

Article
https://doi.org/10.1016/j.clinbiomech.2026.106775
cites this dataset

Licence