Open Muscle

OpenMuscle builds open-hardware forearm force-myography (FMG) bands that read finger motion from the pressure of the forearm muscles, and publishes the recordings so others can train and test models on them. The data is meant for researchers in human-machine interfaces, prosthetic sensors and VR interfaces. The targets are continuous finger flexion and extension, one regression target per frame, not gesture classes.

Our mission is to research, develop, and promote open-source and open-hardware technologies in the biomedical field, with a focus on prosthetic sensors, to enhance accessibility and improve the quality of life for individuals with disabilities.

Hardware

• FlexGrid V4 band: a 15x4 Velostat pressure grid (60 cells) on a flexible PCB, read by an ESP32-S3 at roughly 20 Hz (17.0 to 19.2 Hz measured per band in the dataset), with a 6-axis IMU. It streams over Wi-Fi to a PC.
• LASK5 labeler: a handheld, spring-loaded four-piston finger labeler that records finger position as ground truth while the band records pressure.
• Licenses: hardware CERN-OHL-S-2.0, software MIT, data CC BY 4.0.

Datasets

openmuscle/openmuscle-fmg

• 75 captures, 348,378 sensor frames (603.9 minutes) in four configs: LASK5 labeled (4 piston channels, recorded on the earlier V3 band), Quest hand tracking labeled (25 joints, V3 and V4 bands), a 2025 legacy set, and unlabeled long-wear streams.
• Two wearers (P01, P02). Use temporal or leave-one-session-out splits: random row splits overstate within-session skill by about 2.6x.

Paper

OpenMuscle technical report v0.2: forearm force-myography with VR hand-tracking ground truth, dataset and honest baselines.

Code and hardware

All repositories are under github.com/Open-Muscle.

• OpenMuscle-Hub: project hub and documentation
• OpenMuscle-FlexGrid: FlexGrid band hardware (KiCad)
• OpenMuscle-LASK5: LASK5 labeler hardware
• OpenMuscle-Software: capture, training and analysis software
• FlexGridV4-Firmware: MicroPython firmware for the V4 band
• OpenMuscle-Dataset: the first public data drop (raw CSV, CC BY 4.0)

Get involved

• Questions, ideas and bugs: open an issue on the repository that owns the work, or on OpenMuscle-Hub if you are not sure which one, and a maintainer will route it.
• openmuscle.org/get-involved: wear the hardware to generate diverse training data, write code, or spread the word.

Links

• Website: openmuscle.org
• GitHub: Open-Muscle
• YouTube: @TURFPTAx
• X: @TURFPTAx