A Lightweight Solution for Real-Time Indian Sign Language Recognition

Recognition of signs is one of the fields that is experiencing the fastest growth in research language. Recently, new techniques have emerged in this field. The main form of communication for those who are deaf and dumb is sign language. For those who struggle with speech and hearing, learning real-time sign language is essential. Sign language uses hand gestures for nonverbal communication language. Awareness of these people's language is crucial. In this study, we suggest a novel approach for implementation. The model will provide sign language transcripts for disabled individuals. A live meeting or video conference can replace an in-person session. Sign language problems and applications are currently being tackled with deep learning. The paper seeks to identify Indian sign language utilizing YOLO. Develop a custom YOLO model for sign language recognition. The dataset utilized in this study is downloaded from the Roboflow website and used to train and test the data. In the experiments, YOLOv8 outperformed previous YOLO versions. Based on testing results, YOLOv7 has a greater recall value than YOLOv8, according to precision and mAP measures. Proposed approach shows good result than the other approaches. The model, trained on the Indian sign language letters and numbers dataset, is an efficient choice. Proposed approach shows good precision, recall and [email protected] values 98.46, 98.78, 97.54 than the other approaches.