DATA COLLECTION OF HAND GESTURES ON A HORIZONTALSURFACE USING MEDIAPIPE LIBRARY

Abstract

The horizontal hand gesture recognition is an innovative, cheaper way for human-computer interaction. Currently, most researchers work with sensors, devices for hand gesture recognition, which require more resources. Instead, the presented horizontal method for hand gesture signal recognition by frames, with trained model algorithms. A key element of this work is the research
of a recognition algorithm using only a single camera and collecting dataset to
train a hand recognition model. In the presented framework, the hand detection works with computer vision (CV) algorithms, in general MediaPipe as a converting blue, green, red (BGR) image to red, green, blue (RGB) before processing. There are handedness and hand landmarks on the image as a result of a hand detection. Each point of the landmark has coordination x, y, z values.
The collected dataset will be used to train a model with machine learning (ML)
or neural network algorithms to develop this project as a hand gesture
recognition project. CV, ML, MediaPipe, neural networks, hand gesture, BGR, RGB, human-computer interaction