Title: Study on hand gesture recognition with CNN-based deep learning
Authors: Buemjun Kim; Kyounghee Lee
Addresses: Department of Computer Engineering, Pai Chai University, Daejeon, South Korea ' Department of Computer Engineering, Pai Chai University, Daejeon, South Korea
Abstract: Natural user interface technology is actively studied to enable a computer to understand a human's natural behaviours including gestures and expressions. To recognise human motions, while most approaches generally require special devices such as infrared cameras or motion sensors, we propose a system based on deep learning to recognise a user's hand gestures in general images. A key feature of our system is pre-processing of input images to improve training efficiency and inference accuracy of a deep neural network model to classify hand gesture images. It performs black-white binarisation of each image to effectively distinguish a hand area before training a deep neural network. Our implementation shows the proposed system has a practicality for correctly classifying various hand gesture images such as decimal number 0~9 and Korean alphabet consonants. It is also shown that the confidence degree of those classifications can be considerably improved by the proposed image pre-processing.
Keywords: hand gesture recognition; convolutional neural network; deep learning; image binarisation; natural user interface.
DOI: 10.1504/IJCVR.2021.118532
International Journal of Computational Vision and Robotics, 2021 Vol.11 No.6, pp.571 - 579
Received: 06 Feb 2020
Accepted: 22 May 2020
Published online: 28 Oct 2021 *