Developing Sign Language Recognition Model for Afaan Oromoo Words Using a Deep Learning Techniques

Abstract The 2021 WHO report highlighted that over 5% of the global population experiences hearing impairment, with Ethiopia facing an even higher prevalence of over 10%. Unfortunately, the educational landscape in Ethiopia has not adequately accommodated deaf students, particularly those from the Oromoo nation, as sign language support remains scarce. This lack of accessibility hinders deaf individuals from participating fully in educational opportunities, perpetuating inequality. In response to these challenges, a researcher has proposed an innovative solution utilizing deep learning techniques for automatic recognition of Afaan Oromoo sign language. This model aims to bridge the gap by providing deaf individuals with equal access to education. The proposed system comprises crucial stages including image preprocessing, feature extraction, feature learning, and classification. During preprocessing, video data is converted to image frames, standardized in size, and noise is removed. Feature extraction utilizes Gabor filtering to extract representative features, while convolutional neural networks facilitate feature learning. The model, trained on a dataset of 2025 images, achieved a remarkable 92.98% testing accuracy using ResNet-50 for classification. The researcher emphasizes the importance of further exploration, particularly in handling real-time data with complex background variations. This pioneering effort not only addresses the educational disparity faced by the deaf community in Ethiopia but also sets a precedent for leveraging technology to foster inclusivity and accessibility on a global scale.