EMG-Based hand gesture recognition for myoelectric prosthetic hand control
No Thumbnail Available
Date
2021
Journal Title
Journal ISSN
Volume Title
Publisher
University of Oum El Bouaghi
Abstract
The work presented in this paper aims to contribute to the development of a deep learning-based approach to recognize hand movements using surface electromyography signals (sEMG). This will be used subsequently for the incarnation of these movements using a prosthetic hand.
The developed classification mechanism combines the proprieties of the convolutional neural network (CNN) known by its ability to extract EMG signal features with the properties of the Long short-term memory (LSTM) cells that extract and learn sequential properties of the time series phenomena. The developed approach could achieve an accuracy rate of 98.8% using the "sEMG for Basic Hand movements Data Set" dataset. To our knowledge, the obtained result is one of the highest accuracy values among all researches done using this dataset. To develop a complete prosthetic hand system, a prosthetic hand prototype was designed. It was controlled by the hand recognition unit in order to embody the six movements (cylindrical, hook,
lateral, palmar, spherical, and peripheral) available in the dataset used.
Description
Keywords
Hand gesture recognition; sEMG signals; Deep learning; Convolutional neural
networks (CNN); Long Short- Term Memory (LSTM); Prosthetic hand