Publication: Dynamic hand gesture recognition using deep learning technique
Loading...
Date
2023-08
Authors
Tan, Qian Hui
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Nowadays, the traditional input devices such as mouse, keyboards and remotes are gradually replaced by alternative methods due to the lack of flexibility. Typically, the popular ways for humans to interact with computers include voice commands and body language which are commonly used in commercial electronic products. Hand gestures are the most effective methods of meaningful expression compared to gestures of other body parts. However, there are limitations in detection due to background complexity, illumination variation and occlusion in a vision-based hand gesture recognition system. Complex articulated shape of the hand increases difficulty in modelling the appearance of the hand. In some cases, sub-gesture problem occurs when a gesture is same as a sub-part of a longer gesture. Apart than that, the differentiation between the meaningful and meaningless motion trajectory is a challenging issue for dynamic hand gesture recognition. Hence, a real-time hand tracking and gesture recognition system is proposed. In this project, a real-time hand tracking and landmarks estimation is implemented in PyCharm with the aid of OpenCV and MediaPipe. Then, hand gesture recognition is performed using basic CNN algorithm. The network is built and trained using Keras.