Static Sign Language Word-Level Detection and Recognition for the Oromo Language

Sign language is a way to communicate ideas and feelings to those who are hard of hearing. In this study, we suggested using AI to help hearing-impaired people communicate better. The study depends on the sign language word level of the Afaan Oromo text. The goal of this paper is to use a deep learning strategy to generate static word-level translations from signed words into equivalent Afaan Oromo texts. Afaan Oromo text is the system's final output, and video frames containing text in signed language serve as the system's input. Our study offers a thorough understanding of how YOLO-v9 functions and outperforms the earlier model. We collected literature, conducted an experiment, and used video data. Pre-processing tasks such as frame extraction, resizing, labeling, and data splitting using Roboflow are carried out in order to train our model. The system achieved a precision of 88.8%, a recall of 91.3%, an mAP of 92.7% at 0.5 IoU, and a score of 75.2% at 0.5:0.95 IoU. In general, our model is usable for our community, who can read Afaan Oromo texts, and the visually impaired to recognize Afaan Oromo, because many people cannot hear and understand the signs.