Adaptive Gesture Recognition System for Robotic Control using Surface EMG Sensors

Benjamin Marchiex, Bryan Gardiner, Sonya Coleman

Research output: Contribution to conferencePaperpeer-review

176 Downloads (Pure)

Abstract

Traditionally, Electromyography (EMG) technology was used primarily in the medical domain for the investigation of neuromuscular normalities. However, the development of cheap surface EMG armbands have made this high-end technology commonly accessible to a much wider community. For example, providing gesture interfaces for gaming or controlling peripheral hardware devices. So far, research within this field has typically used complex machine learning classifiers, substantially large databases and long learning/training phases to develop applications-based approaches.
In this paper, a novel algorithm is presented based on one shot learning, i.e. requiring only one example per gesture, therefore substantially reducing the training time and database size. To assess the reliability and the usefulness of the developed system, the accuracy of the algorithm has been compared with classic machine learning approaches providing comparable accuracy. Additionally, the algorithm is successfully demonstrated via a robotic control experiment using various gestures for mobile platform and manipulator control.
Original languageEnglish
Publication statusAccepted/In press - 2 Nov 2019
EventIEEE International Symposium on Signal Processing and Information Technology - Ajman, UAE, Ajman, United Arab Emirates
Duration: 10 Dec 201912 Dec 2019

Conference

ConferenceIEEE International Symposium on Signal Processing and Information Technology
Abbreviated titleISSPIT
Country/TerritoryUnited Arab Emirates
CityAjman
Period10/12/1912/12/19

Keywords

  • Electromyography (EMG)
  • robotic control
  • adaptive gesture recognition

Fingerprint

Dive into the research topics of 'Adaptive Gesture Recognition System for Robotic Control using Surface EMG Sensors'. Together they form a unique fingerprint.

Cite this