Home / Current Issue / Paper 1708316
Gesture Recognition Using MediaPipe and Keras for Smart Interfaces
Subject area: Science,Engineering and Technology · Area of research: Engineering
Abstract
It creates a significant distance between user and the computer, the use of a physical controller, such as a mouse or keyboard, hinders a natural interface in human-computer interaction. We have developed a reliable marker-less hand gesture recognition system in this paper that effectively tracks both dynamic and static hand gestures. When a gesture is detected, our system converts it into actions like opening websites and starting programs like PowerPoint and VLC Player. In a presentation, the gesture is used to move through the slides. Our findings demonstrate that minimal hardware requirements and can be used to create an intuitive HCI. Despite being a highly regarded technique for human-computer interaction, hand gesture recognition still faces many obstacles. The issue of individual user style variation, which has a substantial impact on system performance, is addressed in this paper. An efficient, flexible graphical user interface that supports user-defined hand gestures is presented here in this project. In contrast, previous research only permitted the manual addition of personalized hand gestures in extremely limited application contexts. Our system uses data from a specific user to train a camera-based hand gesture recognition. model for that user, allowing hand gestures to be personalized. In contrast to earlier recognition models that require large training datasets, we use a lightweight Multilayer Perceptron architecture based on contrastive learning, which reduces the size of the data required and the training timeframes. While a user study gathers and examines some preliminary user feedback on the system in use, experimental results show that the recognition model converges quickly and achieves satisfactory accuracy.
How to cite this paper
@article{1708316,
author = {Aniket Patel, Harshit Katiyar, Kushal Gupta},
title = {Gesture Recognition Using MediaPipe and Keras for Smart Interfaces},
journal = {Iconic Research And Engineering Journals},
year = {2025},
volume = {8},
number = {11},
pages = {1992-1996},
issn = {2456-8880},
url = {https://www.irejournals.com/formatedpaper/1708316.pdf},
abstract = {It creates a significant distance between user and the computer, the use of a physical controller, such as a mouse or keyboard, hinders a natural interface in human-computer interaction. We have developed a reliable marker-less hand gesture recognition system in this paper that effectively tracks both dynamic and static hand gestures. When a gesture is detected, our system converts it into actions like opening websites and starting programs like PowerPoint and VLC Player. In a presentation, the gesture is used to move through the slides. Our findings demonstrate that minimal hardware requirements and can be used to create an intuitive HCI. Despite being a highly regarded technique for human-computer interaction, hand gesture recognition still faces many obstacles. The issue of individual user style variation, which has a substantial impact on system performance, is addressed in this paper. An efficient, flexible graphical user interface that supports user-defined hand gestures is presented here in this project. In contrast, previous research only permitted the manual addition of personalized hand gestures in extremely limited application contexts. Our system uses data from a specific user to train a camera-based hand gesture recognition. model for that user, allowing hand gestures to be personalized. In contrast to earlier recognition models that require large training datasets, we use a lightweight Multilayer Perceptron architecture based on contrastive learning, which reduces the size of the data required and the training timeframes. While a user study gathers and examines some preliminary user feedback on the system in use, experimental results show that the recognition model converges quickly and achieves satisfactory accuracy.},
month = {May},
}