Real-Time Sign Language Recognition using MediaPipe and 1D CNN

Abstract

Sign language recognition is a critical technology that enhances communication for individuals who are deaf or hard of hearing. This study presents a real-time sign language recognition system that combines MediaPipe for efficient feature extraction and a 1D Convolutional Neural Network (CNN) for accurate gesture classification. Our approach focuses on dynamic American Sign Language (ASL) signs and addresses key challenges in the field, including real-time performance, hardware cost, and adaptability. The dataset comprises 15 dynamic ASL signs, recorded under various lighting conditions and backgrounds to ensure robustness. The proposed model achieves 98% accuracy in recognising the 15 dynamic ASL signs. Notably, our system operates in real time with only a standard webcam, making it both affordable and easily deployable.