Gesture Learn: A Real-Time CNN-Based Hand Gesture Recognition System Using MediaPipe
Year: 2026
Type: conference-paper
Abstract: In the fast-growing sphere of human-computer interaction, gesture input performed by means of hands has emerged as a natural and efficient way of conducting touchless communication. This paper proposes Gesture Learn, an AI-based platform that can decode sign language alphabets in real-time on uninterrupted camera input. The system focuses on finding significant visual features reflecting the patterns of hand gestures that are static and used in sign language. Gesture Learn integrates the current computer vision pipeline with the MediaPipe framework, which extracts hand landmarks accurately, and employs a Convolutional Neural Network (CNN) to reinforce the accuracy of gesture detection. Training was performed by generating a custom dataset by recording live gestures and then adopting a systematic process for training and validating the model. The overall design is user-friendly, responsive, and provides smooth real-time functionality. Experimental evaluation demonstrates high and stable accuracy and consistency of results.
Institutions: Galgotias University, Manipal Academy of Higher Education
Language: English
Subfield: Human-Computer Interaction
Field: Computer Science
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W7171531328