Real Time Hand Gesture Recognition Using Leap Motion Controller Based on CNN-SVM Architechture

Aamrah Ikram, Yue Liu

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

9 Citations (Scopus)

Abstract

In rapidly growing field of Artificial Intelligence (AI), Hand Gesture Recognition (HGR) is an important entity. In the real world system it is very challenging to detect and classify Dynamic Hand Gestures (DHG). As there is considerable diversity in gesture performed by individuals and the system should be real time to overcome the delay between performing and classifying the gesture. In this work, we proposed a new approach for efficient HGR using Convolutional Neural Network (CNN) along with Support Vector Machine (SVM) classifier. CNN used to avoid feature extraction and to minimized the number of trained parameters. However, to reduce the error, Error Break Propagation Algorithm (EBPA) is implemented. For the system's validity and robustness SVM optimizer has been used. An overall accuracy of 93 % has achieved on DHG 14/28 dataset.

Original languageEnglish
Title of host publication2021 IEEE 7th International Conference on Virtual Reality, ICVR 2021
PublisherInstitute of Electrical and Electronics Engineers Inc.
Pages5-9
Number of pages5
ISBN (Electronic)9781665423090
DOIs
Publication statusPublished - 20 May 2021
Externally publishedYes
Event7th IEEE International Conference on Virtual Reality, ICVR 2021 - Foshan, China
Duration: 20 May 202122 May 2021

Publication series

NameInternational Conference on Virtual Rehabilitation, ICVR
Volume2021-May
ISSN (Electronic)2331-9569

Conference

Conference7th IEEE International Conference on Virtual Reality, ICVR 2021
Country/TerritoryChina
CityFoshan
Period20/05/2122/05/21

Keywords

  • convolution neural network
  • dynamic hand gesture
  • support vector machine
  • virtual reality

Fingerprint

Dive into the research topics of 'Real Time Hand Gesture Recognition Using Leap Motion Controller Based on CNN-SVM Architechture'. Together they form a unique fingerprint.

Cite this