Computer vision with Microsoft Kinect for control of functional electrical stimulation: ANN classification of the grasping intentions

Matija D. Štrbac, Dejan B. Popović

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

1 Citation (Scopus)

Abstract

We present a method for recognizing intended grasp type based on data from the Microsoft Kinect. A computer vision algorithm estimates the vertical and the transversal distance of the hand from the center of the object and the hand orientation from the Kinect depth images. Based on this set of features in the reaching phase of grasp artificial neural network recognizes the intended grasp type. This is demonstrated with an example of a coffee cup on a working desk. Trained neural network classified the grasp with accuracy above 85%. By adding this feature to the existing computer vision system for control of the functional electrical stimulation assisted grasping we facilitate the compliance between the applied electrical stimulation and the user intentions.

Original languageEnglish
Title of host publication12th Symposium on Neural Network Applications in Electrical Engineering, NEUREL 2014 - Proceedings
EditorsBranimir Reljin, Srdan Stankovic
PublisherInstitute of Electrical and Electronics Engineers Inc.
Pages153-156
Number of pages4
ISBN (Electronic)9781479958887
DOIs
Publication statusPublished - 15 Jan 2014
Event12th Symposium on Neural Network Applications in Electrical Engineering, NEUREL 2014 - Belgrade, Serbia
Duration: 25 Nov 201427 Nov 2014

Publication series

Name12th Symposium on Neural Network Applications in Electrical Engineering, NEUREL 2014 - Proceedings

Conference

Conference12th Symposium on Neural Network Applications in Electrical Engineering, NEUREL 2014
Country/TerritorySerbia
CityBelgrade
Period25/11/1427/11/14

Keywords

  • Computer vision
  • FES
  • Microsoft Kinect
  • neural networks
  • rehabilitation

Fingerprint

Dive into the research topics of 'Computer vision with Microsoft Kinect for control of functional electrical stimulation: ANN classification of the grasping intentions'. Together they form a unique fingerprint.

Cite this