IMoveLab

Guided tutorial

Applying IMoveLab with Claude

Install Claude Code first

Use the official installer, confirm it works, then sign in.

Official quickstart

Install the CLI

Open a terminal and run the command for your system.

macOS, Linux, or WSL curl -fsSL https://claude.ai/install.sh | bash

Windows PowerShell irm https://claude.ai/install.ps1 | iex
Verify the install

Confirm that Claude Code is available.
claude --version
Sign in

Start Claude Code and follow the browser prompts.
claude

Once Claude Code is installed, follow this 6-minute tutorial to set up and run IMoveLab locally.

Experiment 1

Accuracy Evaluation

19Healthy athletes

4Dynamic tasks

3Sensing modalities

Experiment 1 quantifies the accuracy of IMU-based kinematic tracking against a high-fidelity bone-level reference. Nineteen healthy athletes performed four dynamic tasks while three sensing modalities recorded their movement simultaneously.

Tasks: running, single-leg drop, double-leg drop, and single-leg hop.
Modalities: inertial measurement unit (IMU), marker-based motion capture, and biplane radiography — the ground-truth reference.

Experiments 2 & 3

Drift Evaluation

Experiment 2

Minute-long activities

23Adults

11Tasks

2Modalities

Tasks: 11 minute-long tasks — 3 locomotion tasks and 8 physical-therapy exercises.
Modalities: inertial measurement unit (IMU) and marker-based motion capture (reference).

Experiment 3

Hour-long trial

5Adults (subset of Exp. 2)

~1 hrContinuous trial

2Modalities

Protocol: three 10-minute walking trials separated by two 20-minute resting periods.
IMU: continuous capture throughout the full hour.
Motion capture: reference recording during the walking trials only, turned off during resting.

Citation

Cite this work

If you use any of the data or code, please cite our paper:

Phan, V., Li, Z., Meinders, E. et al. “Biomechanics-informed inertial tracking achieves the accuracy of marker-based kinematics.” Nature Communications (2026). DOI: https://doi.org/10.1038/s41467-026-75981-y

Acknowledgements

This work was funded by the National Science Foundation CAREER Award (CBET 2145473) and the Dowd Fellowship from the College of Engineering at Carnegie Mellon University. We would like to thank the volunteers for their time participating in this study.

Contact