Title:Developing Lightweight DNN Models With Limited Data For Real-Time Sign Language Recognition

Abstract:We present a novel framework for real-time sign language recognition using lightweight DNNs trained on limited data. Our system addresses key challenges in sign language recognition, including data scarcity, high computational costs, and discrepancies in frame rates between training and inference environments. By encoding sign language specific parameters, such as handshape, palm orientation, movement, and location into vectorized inputs, and leveraging MediaPipe for landmark extraction, we achieve highly separable input data representations. Our DNN architecture, optimized for sub 10MB deployment, enables accurate classification of 343 signs with less than 10ms latency on edge devices. The data annotation platform 'slait data' facilitates structured labeling and vector extraction. Our model achieved 92% accuracy in isolated sign recognition and has been integrated into the 'slait ai' web application, where it demonstrates stable inference.