Title: ANN-based EMG classification for myoelectric control

Authors: Rami J. Oweis; Remal Rihani; Afnan Alkhawaja

Addresses: Faculty of Engineering, Biomedical Engineering Department, Jordan University of Science and Technology, P.O. Box 3030, Irbid 22110, Jordan ' Faculty of Engineering, Biomedical Engineering Department, Jordan University of Science and Technology, P.O. Box 3030, Irbid 22110, Jordan ' Faculty of Engineering, Biomedical Engineering Department, Jordan University of Science and Technology, P.O. Box 3030, Irbid 22110, Jordan

Abstract: This work presents a new neural network model related to EMG signal classification for myoelectric control. The aim of this work is to develop a more accurate method for pattern recognition and intention interpretation of five human forearm hand gestures: grasping, extension/flexion, and ulna/radial deviation. A sum of 750 signals that incorporated all the selected hand movements were acquired from five volunteers, preprocessed, and then time and time-series domain features were extracted. Classification model in MATLAB platform is then utilised for classification purposes. The neural network classifier achieved an average accuracy up to 96.7%. The system overall average validation parameters calculated for the five movements were: sensitivity of 96.9%, specificity of 99.0%, PPV of 96.9%, and NPV = 99.1%.

Keywords: EMG dignals; EMG classification; electromyography; myoelectric control; artificial neural networks; ANNs; feature extraction; classification modelling; pattern recognition; intention interpretation; forearm hand gestures; hand movements.

DOI: 10.1504/IJMEI.2014.065442

International Journal of Medical Engineering and Informatics, 2014 Vol.6 No.4, pp.365 - 380

Received: 03 Oct 2013
Accepted: 28 Feb 2014
Published online: 31 Oct 2014 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article