| 한글제목(Korean Title) |
Study on Recognition of Hand Gestures Using Convolutional Neural Network |
| 영문제목(English Title) |
Study on Recognition of Hand Gestures Using Convolutional Neural Network |
| 저자(Author) |
Buemjun Kim
Kyounghee Lee
|
| 원문수록처(Citation) |
VOL 11 NO. 01 PP. 0236 ~ 0239 (2019. 06) |
한글내용
(Korean Abstract) |
|
영문내용
(English Abstract) |
Currently, natural user interface technology is actively studied to enable users to communicate with a computer by understanding their ordinary behaviors and expressions. A lot of those researches focus on recognition of human's motions such as a hand gesture. While most existing motion recognition approaches require special devices such as an infrared camera or a motion sensor, this paper proposes a system employing deep learning technology to recognize a user's hand gesture from an image in normal real-time video produced by general devices such as a webcam. To enhance the recognition accuracy through deep learning, the proposed system first performs black-white binarization processing of each image to effectively distinguish an area corresponding to a user's hand. Then those images are used for learning and inference of our convolutional neural network model to distinguish various hand gestures. With our implementation, we performed experiments to recognize some simple hand gestures representing rock-paper-scissors. After sufficient learning with more than 100 images per each gesture, the proposed system could accurately infer meanings of all test images and the degree of confidence was higher than 90%. Our further study could be extending the proposed system to recognize more complex hand gestures such as decimal numbers and alphabets. Finally, it is expected to understand in realtime various connected hand motions such as a sign language.
|
| 키워드(Keyword) |
Convolutional Neural Network
Deep Learning
Hand Gesture Recognition
Natural User Interface
|
| 파일첨부 |
PDF 다운로드
|
