Surface EMG Feature Dataset for Hand Movement Classification

Abstract

Electromyography is a bioelectric signal recorded during muscle contraction or relaxation. The EMG signal is subject‐specific, reflecting which muscle is contracting at what intensity. In recent years, measuring electromyography has become easier with measuring devices designed as bracelets. Previous studies focused on classifying gestures of the entire hand or wrist, as gestures that only use the fingers, such as gripping an object, are often more challenging to classify. Data acquisition was performed using the MindRove Armband EMG, a wearable device consisting of eight sensors arranged in a bracelet‐like configuration. The EMG dataset was collected from 15 subjects performing the movement of grasping eight different objects, repeated five times. The purpose of the study is to propose a surface EMG feature dataset for hand movement classification. Four time‐domain features were extracted and used as inputs for classification models, including artificial neural networks (ANN), K‐nearest neighbors (KNN), and Decision Trees. The ANN model achieved the highest classification accuracy of 93.72%. Overall, the classification results suggest that the proposed dataset provides reliable sEMG signals and is suitable for EMG‐based hand gesture recognition research. © 2026 Institute of Electrical Engineers of Japan and Wiley Periodicals LLC.