Datasets:

Modalities:
Video
Size:
< 1K
Libraries:
Datasets
License:
LeRobot Worldwide Hackathon org
No description provided.
LeRobot Worldwide Hackathon org

🦾 EMG Hand Gesture Dataset — 3-Channel MyoWare Signals
This dataset was collected as part of the “AI Prosthetic Hand Control via the Peripheral Nervous System”
📦 Dataset Summary
This dataset contains 3-channel surface EMG (sEMG) signals recorded from the forearms of human participants while performing hand gestures such as "Rest", "Fist", "Paper", and "Okay". The data was collected using MyoWare 2.0 sensors placed on:

Brachioradialis

Flexor Carpi Ulnaris

Flexor Carpi Radialis

The signals were sampled at 1070 Hz and segmented into 300 ms overlapping windows (stride: 30 ms), following best practices for real-time gesture recognition.

Each episode contains:

A full window of 3-channel EMG signals

The corresponding gesture label

Accurate time metadata

Subject and trial metadata

The data is stored in the LeRobotDataset format for compatibility with robotics and imitation learning workflows.

🧠 Project Context
This dataset supports the development of affordable AI-based prosthetic hands using pattern recognition. The prosthetic hand is 3D-printed and controlled using Raspberry Pi, servo motors, and a trained AI model (ANN).

The EMG signals enable gesture classification, and the predicted output is used to control the fingers of the prosthetic hand in real time.

👥 Participants
8 participants: 4 male and 4 female

Each participant performed 4 gestures: Rest, Fist, Paper, Okay

Each gesture was repeated across 4 rounds, with 5 repetitions per round

🏷 Labels
Rest

Fist

Paper

Okay

Each episode is labeled with the most frequent gesture in its 300 ms window.

imstevenpmwork changed pull request status to merged

Sign up or log in to comment