sEMG-MANUS dataset: synchronised Myo sEMG and MANUS finger joint recordings for hand pose estimation and XR musical interaction
Summary
No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.
Summarise with VIRAT Free to try with your Science Catalogs account.Abstract
sEMG-MANUS Dataset This repository is the public companion repository for a multimodal dataset of synchronised surface electromyography (sEMG) and finger joint angle recordings captured with a Thalmic Labs Myo armband and a MANUS Quantum Metaglove. The dataset was created to support research on data-driven hand pose estimation, gesture analysis, and XR musical interaction. The primary methodological description appears in Chapter 8 of my PhD thesis _When XR Meets AI: Design and Machine Learning for Performer-Centred Gestural Control in Extended Reality Musical Instruments_. Records - Dataset DOI: `10.5281/zenodo.19261324`- Recording application: [`sEMG-manus-manager`](https://github.com/maxgraf96/sEMG-manus-manager) What Is Included Here - publication metadata for GitHub and Zenodo- dataset documentation and codebook- data-quality notes for known missing and duplicate trials- local scripts to build a Zenodo-ready package from the ignored `data/` tree Dataset Scope - Modalities: - 8-channel Myo sEMG - 10 Myo IMU channels - 20 MANUS finger-joint channels - 4 wrist quaternion channels- On-disk CSV schema: 42 columns per sample- Gesture taxonomy: 22 published gestures- Participant folders in the full release: 18- Recommended benchmark cohort: 15 participants (`u_3` to `u_16` and `u_18`) Cohort Note The published Zenodo package contains 18 participant folders for completeness. However, `u_1`, `u_2`, and `u_17` are incomplete and should be excluded from balanced between-user analyses unless a study explicitly models missingness or partial data. The 15-participant benchmark cohort is therefore: - `u_3` to `u_16`- `u_18` Even within that core cohort, the raw tree contains a small number of trial-count anomalies: - `u_7/s_1/g_thumbs_up` has 2 recordings instead of 3- `u_8/s_1/g_tap_middle` has 2 recordings instead of 3- `u_9/s_2/g_tap_pinky` has 4 recordings instead of 3 The local QC scripts and generated manifest flag these cases explicitly. File Layout In The Archive The data included here contains: - `semg-manus-dataset-v1.zip`- `README.md`- `CODEBOOK.md`- `DATA_QUALITY.md`- `manifest.csv`- `checksums.txt`- `LICENSE-data-CC-BY-4.0.txt` Inside `semg-manus-dataset-v1.zip`, the raw tree is preserved as: ```textdata/ u_/ s_/ g_/ recording__ _____.csv``` Citation If you use this dataset, cite both the Zenodo dataset record and the companion repository/software release once the DOIs have been assigned. Suggested citation text: > Graf, Max, and Mathieu Barthet. sEMG-MANUS dataset and companion repository. Queen Mary University of London. Dataset DOI: `10.5281/zenodo.19261324`.
Fields of Study & Subjects
Hand Gesture Recognition Systems
Muscle activation and electromyography studies
Human Motion and Animation
Publication details
- Journal / source
- Zenodo (CERN European Organization for Nuclear Research)
- Publication date
- 27 March 2026
- Language
- EN
- Document type
- Dataset