Hand Gesture Recognition using CNN-GNN

This research study proposes an innovative hand gesture recognition system designed to enhance human-computer interaction (HCI) through natural communication. The system operates in three stages: Learning, Detection, and Recognition. In the Learning phase, a training dataset of various hand gestures is utilized for feature extraction, focusing on determining centroids to geometrically divide images. The Detection phase captures real-time images via a webcam, employing histogram clustering to identify hand regions. Finally, the Recognition phase integrates Convolutional Neural Networks (CNN) and Graph-based Neural Networks (GNN) to accurately recognize gestures by analyzing defect points and understanding relational dependencies between gestures and context. Experimental validation demonstrates the system’s efficacy in interpreting gestures for actions like page switching and scrolling. Through a comprehensive literature review, the study highlights limitations in existing methods and presents a novel architecture that incorporates spatial and temporal modeling, significantly improving gesture recognition accuracy and real-time performance. By leveraging advanced machine learning algorithms and deep neural networks, this system provides a robust solution for creating user-friendly interfaces and efficient human-machine interactions.