Classification of Hand Movements Using Electromyographic Signals and Machine Learning Models

In this work, the challenge of classifying three hand movements from electromyographic (EMG) signals is addressed. Eight fundamental features, such as peak value, integrated EMG, mean absolute value, and others, were extracted from the acquired EMG signals. Five classifier models were evaluated: Support Vector Machine (SVM), Decision Tree, Random Forest, Radius Neighbors and Gradient Boosting. The results showed that the classifiers based on decision trees obtained high precision (97%), standing out in the classification of movements. However, it will be noted that the SVM and Radius Neighbors will need additional tuning to improve their performance (61% and 84% accuracy, respectively). The extracted features are shown to be relevant for classification, providing valuable information to differentiate between classes. This study emphasizes the importance of selecting appropriate models and suitable features to improve the accuracy of EMG signal classification in future applications.