Implementation and Evaluation of a Real Time Sign Language Recognition System
Ganesh S. Khekare, Pranay Batta, Tavishi Jain, Ajay Kumar Phulre, Vipin Singh, Goutam Majumder · 2025
This research presents the deployment and assessment of a real-time sign language identification technique with 87% accuracy for gesture and sound recognition and 450ms latency. Using a webcam interface, the system utilizes an adaptive ensemble learning architecture of CNN, RNN, and Transformer models for static and dynamic gesture classification. Environmental testing in adverse lighting and background noise environments consistently performs (>85% accuracy). Novel features include cross-platform compatibility with desktop and mobile platforms, bidirectional communication with sign-to-text/speech and speech-to-sign translation support, and scalable architecture supporting different signing styles. Preprocessing techniques for improved occlusion and lighting robustness ensure system performance in real-world operation. User testing with deaf and hearing subjects showed promising feedback in terms of usability and communication effectiveness. This inclusive solution bridges substantial communication barriers while offering a platform for future growth in vocabulary extension and regional sign language adaptation.