Surface EMG Classification of Basic Hand Movement

In recent decades, enormous improvements in the rehabilitation of the amputee have been made. A major development is that the prosthetic restoration of a limb or an extremity re-establishes the functional mobility of the patient. Therefore, myoelectric control devices become popular as they mimic the appearance and replace the function of a missing limb simultaneously. For this reason, the overall goal of this project is to classify motion patterns, especially of different types of contractions in order to drive a below-elbow prosthetic arm. Firstly, the acquisition part is designed to capture surface EMG signals using Myoware sensor acquisition board placed on the biceps and triceps and to perform the required signal processing. Secondly, an analysis is performed on the obtained signals, and feature extraction is implemented accordingly. After that, the EMG classification is done using an Artificial Neural Network (ANN) computing system. The ANN is a machine learning method made of many layers, and trained to recognize the muscle contractions. The weights and the activation function are updated to predict seven gestures output. Thus, the input and the output are manipulated to perform a specific task accurately and the algorithm is considered a real time system due to its fast response.