Akademik Veri Yönetim Sistemi
12th International Conference on Automation, Robotics and Applications, ICARA 2026, İstanbul, Türkiye, 5 - 07 Şubat 2026, ss.550-555, (Tam Metin Bildiri)
This study presents a robust methodology for the development of a control system for a bionic hand prosthesis based on electromyography (EMG) signals and machine learning methods. The system is designed to provide intuitive control by interpreting muscle activity collected from the residual limb with surface EMG sensors. The acquired signals are processed using a sliding analysis window and Mean Absolute Value (MAV) features before classification. Several machine learning algorithms, including Linear Discriminant Analysis (LDA), Support Vector Machines (SVM), and Multi-Layer Perceptron Neural Networks (NN), were tested. Two independent neural network models were trained for the thumb-open and thumb-closed configurations, allowing the system to correctly classify a total of eight different hand movements. Both models were deployed on a microcontroller and achieved around 99% accuracy during real-time tests. The system provides stable and responsive control, offering a practical step toward adaptive and user-friendly prosthetic operation.