EMG2Gesture Sample Dataset

EMG2Gesture is a dataset and codebase for surface electromyography (sEMG)-based gesture recognition and prosthetic hand control research. It builds on established low-cost EMG acquisition paradigms for assistive technology and machine learning-based biosignal decoding. The dataset is designed to support research in EMG signal processing, gesture classification, and data-driven prosthetic control systems. Overview This dataset contains surface EMG recordings collected for the development and evaluation of machine learning models for gesture recognition in prosthetic control contexts. It follows a similar experimental structure to prior low-cost EMG prosthetic datasets, where signals are recorded under constrained electrode configurations and across multiple sessions to capture variability in EMG signals over time. The dataset supports reproducible experimentation in EMG-based classification and signal alignment workflows. Experimental Setup EMG signals are acquired using a low-cost surface EMG acquisition system Recordings are collected from multiple participants across multiple sessions A sparse electrode configuration is used to simulate realistic prosthetic control constraints The dataset includes multiple hand gestures relevant to prosthetic hand operation This setup introduces variability across: Subjects, Recording sessions, and Gesture executions which are important for evaluating model robustness. Data Structure The dataset includes: Raw surface EMG recordings, Label information corresponding to gesture intervals, and Unsegmented recordings that require alignment and preprocessing. Gesture segmentation is then performed using the provided alignment and preprocessing pipeline. Processing Pipeline The repository provides tools for: Signal filtering and preprocessing Gesture interval alignment Feature extraction from EMG signals Training machine learning models for classification Evaluating model performance across sessions and subjects