Real-Time Classification of Multi-Channel Forearm EMG to Recognize Hand Movements using Effective Feature Combination and LDA Classifier

  • Muhammad Sabbir Alam Department of Biomedical Physics and Technology, University of Dhaka, Dhaka - 1000
  • ASM Shamsul Arefin Department of Biomedical Physics and Technology, University of Dhaka, Dhaka - 1000
Keywords: EMG, Hand movements, Linear Discriminant Analysis (LDA), Principal Component Analysis (PCA)

Abstract

Electromyography (EMG) signals acquired from surface of arms can be crucial in recognizing nature of hand gestures. The concept is used in current highly demanding fields such as controlling prosthetic limbs, diagnosing neuromuscular disorders, manipulation of robotic arm etc. The purpose of the work was to classify a set of hand motions from corresponding multi-channel surface EMG signals by developing MATLAB tools. The research focused on extracting multiple signal features and finding the appropriate combination of extracted intelligible features to get the best classification accuracy for the specific set of hand gestures. For dynamic and fast classification purpose, linear discriminant analysis (LDA) classifier was employed. Effect of feature dimensionality reduction on classification accuracy was also investigated via Principal Component Analysis (PCA) in this research. Finally, the research analyzed different electrode placements by comparing classification accuracy for each of the set of motions and proposed a simple and compact data acquisition instrumentation having less number of electrodes while maintaining high classification accuracy.

Bangladesh Journal of Medical Physics Vol.10 No.1 2017 25-39

Downloads

Download data is not yet available.
Abstract
107
PDF
55
Published
2018-12-03
How to Cite
Alam, M. S., & Arefin, A. (2018). Real-Time Classification of Multi-Channel Forearm EMG to Recognize Hand Movements using Effective Feature Combination and LDA Classifier. Bangladesh Journal of Medical Physics, 10(1), 25-39. https://doi.org/10.3329/bjmp.v10i1.39148
Section
Original Papers