Developing Human Hand Gesture Data

Abstract

Gesture is a movement originating from one of the body's limbs, generally using the hand. The movement is usually used to interact with each other among human beings. In this study, gestures are considered as a way for humans to be able to communicate with a system so that it can understand human body language. The first step to be able to create a system that can recognize body gestures is to create a dataset of the gestures to be used. The process for creating a dataset in this study includes taking pictures from a webcam, then pre-processing the images so that they can display the characteristics of each image. The dataset can then be used to model gesture classification using Convolutional Neural Network (CNN). The collection of gestures in this dataset is carried out on the entire upper body to create data variations of male and female also with various human body shapes. This study has collected 450 images of gesture data and has prepared these data with ration of 80% for training and 20% for testing.