Biomechanics-Informed Inertial Tracking Achieves the Accuracy of Marker-Based Kinematics

IMoveLab is a workflow for human motion tracking with inertial measurement units (IMUs). A web app for users who may not wish to interact with the code is also being developed and available at IMoveLab.

📂 Structure

See biplane_ref for the implementation associated with Experiment 1 presented in the paper; mocap_ref for Experiments 2 and 3.

⚙️ Installation

The code was tested with Python 3.10.19 on Ubuntu 24.04 and MacOS. We recommend using conda to create a separate environment for running the code.

conda create --name imovelab python==3.10.19
conda activate imovelab

Run python -m pip install -r requirements.txt for dependencies.

See AHRS, VQF, and RIANN regarding packages for state-estimation filters.

See Scripting in Python regarding the opensim package for OpenSense biomechanical constrained inverse kinematics.

💾 Data

The dataset is available on FigShare.

See README.md in the biplane_ref or mocap_ref folder for how to store the data.

🚀 Run the Code

See README.md in the biplane_ref or mocap_ref folder for how to run the code.

🔑 License

IMoveLab is licensed under the Carnegie Mellon Software License.

📝 Citation

Coming soon.

📧 Contact

Please reach out to Vu Phan (vuphan@andrew.cmu.edu) or Eni Halilaj (ehalilaj@andrew.cmu.edu) for any questions regarding this work.