Machine learning and processing techniques for the enhancement of hand gesture recognition of Forcemyography and Electromyography signals

Asfour, Mohammed (2022) Machine learning and processing techniques for the enhancement of hand gesture recognition of Forcemyography and Electromyography signals. Masters thesis, Memorial University of Newfoundland.

[img] [English] PDF - Accepted Version
Available under License - The author retains copyright ownership and moral rights in this thesis. Neither the thesis nor substantial extracts from it may be printed or otherwise reproduced without the author's permission.

Download (2MB)

Abstract

Hand gesture recognition is the primary driver of many applications across user groups. It uses machine learning classifiers to classify users’ data, most prominently Electromyography (EMG) or Forcemyography (FMG). Whereas EMG sensors detect signals going down the arm to the muscles, FMG sensors measure the pressure change on the arm’s skin. Nevertheless, many inconsistencies impact gesture recognition drastically. For instance, gesture recognition for the same user, intra-subject, is affected by the duration between the collected signals for classifiers’ training and the classified signals, requiring more data from the user. A more significant hurdle is inter-subject gesture error, in which classifiers are trained on signals from one or more subjects perform exceptionally poorly on the signals of another. These issues arise due to the uniqueness of such signals per person and their variance through time. We offer methods to encounter several downsides of EMG and FMG. We propose a machine learning pipeline that yields features of consistent performance across various classifier types and reduces intra-subject signal variance. To tackle other intra-subject errors, we offer a ranking for what we define as the feature-classifier compatibility relationship that controls the recognition performance. The methods are tested on FMG and EMG, respectively, and enhanced gesture recognition.

Item Type: Thesis (Masters)
URI: http://research.library.mun.ca/id/eprint/15839
Item ID: 15839
Additional Information: Includes bibliographical references (pages 56-66)
Keywords: hand gestures recognition, machine learning, Electromyography, Forcemyography
Department(s): Science, Faculty of > Computer Science
Date: December 2022
Date Type: Submission
Digital Object Identifier (DOI): https://doi.org/10.48336/9XRF-CE18
Library of Congress Subject Heading: Machine learning; Electromyography; Tactile sensors; Gesture recognition

Actions (login required)

View Item View Item

Downloads

Downloads per month over the past year

View more statistics