RECOGNIZING VIETNAMESE SIGN LANGUAGE

Abstract

The research focuses on employing image recognition technology to address challenges in recognizing Vietnamese sign language into text. Using YOLOv8 (You Only Look Once - version 8) and LSTM (Long ShortTerm Memory) via Mediapipe, this research emphasizes image processing and sign language data. The data undergoes preprocessing and training within the model for recognition, subsequently being converted in real-time into text. Research results demonstrate a significant improvement in sign language recognition by combining YOLOv8 and LSTM via Mediapipe Hands. The model consistently achieves accurate recognition of sign languages and real-time conversion into text. The study showcased the immense potential of utilizing YOLOv8 and LSTM through Mediapipe to support everyday communication for sign language users. However, further efforts are required to optimize and implement these findings in practical applications.