Utilizing Kohonen Maps in Electromyography Recognition System for Upper Limb Prosthetics

This study focuses on evaluating a specific type of artificial neural network known as Kohonen Map (KM) or Self-Organizing Map (SOM), for the characterization of Electromyography datato utilize in upper limb prosthetics. Prosthetic devices aim to assist individuals by restoring some of the functions of a lost limb. Significant research has been dedicated to developing artificial hands that replicate the capabilities of human hands. However, the complexity of the human hand makes it difficult to imitate. Various classifiers have been investigated to enhance the performance of prosthetic hands. In this work, the Kohonen map was utilized on a dataset obtained from a sensor which was built bya local research team called Electronic Prostheses Engineering Team (EPET). This dataset includes 36data sets representing open and closed hand movements from three individuals. Experiments were conducted to classify two hand gesture types open and closed for three individuals, and results displayed as 2D maps indicate that an accuracy exceeding 97% was achievable. Keywords: upper limp prosthesis, Kohonen map, Self Organizing map.