Acquisition and Classification of a Surface EMG Dataset for Finger Movement

Electromyograms (EMGs) are signals generated by changes in the membrane potential of living cells when they are stimulated. There have been many studies using electromyographic (EMG) signals to recognize hand gestures; however, most of them focus on wrist and hand gestures due to more pronounced muscle contractions compared to finger gestures. In this study, we propose a classifier for finger gestures based on an electromyography dataset self‐collected from 10 subjects. The EMG signals, collected from the fixed electrodes of the Muscle BioAmp Patchy sensor from UpSideDown Labs, have six features extracted from their time domain and then used in Artificial Neural Network and K‐Nearest Neighbors classification models. The experimental results show that the highest classification accuracy achieved was 85.4% ± 2.9% for recognizing 10 different finger gestures. Even though only a single signal channel with a simple configuration was used, the system still achieved an acceptable classification accuracy for recognizing 10 different finger gestures, demonstrating the feasibility of EMG‐based finger gesture recognition despite weaker muscle activations compared to wrist and hand gestures. The method we propose provides a cost‐effective solution while ensuring efficiency, making it suitable for deployment in practical applications and reducing computational costs.