Title: Development of Deep Learning Model For Detection of Sign Language Alphabets in an Image
Authors: Oladimeji Olayanju, Adeniji oluwashola david, Adeniyi Michael Odejayi
Volume: 9
Issue: 8
Pages: 18-21
Publication Date: 2025/08/28
Abstract:
Sign language serves as a crucial means of communication for the deaf and hard-of-hearing community. However, the lack of widespread understanding among non-signers presents a communication barrier. This research focuses on the development of a deep learning-based model for the automatic recognition of sign language alphabets in images. The approach in this research leverages Convolutional Neural Networks (CNNs) integrated with Attention Mechanisms, specifically the Convolutional Block Attention Module (CBAM), to enhance feature extraction and improve recognition accuracy. EfficientNet is utilized as the backbone network due to its high performance and parameter efficiency. A dataset of Sign Language gestures is collected from a repository and preprocessed to train and evaluate the model. Experimental results demonstrate the effectiveness of the developed approach in accurately classifying sign language alphabets in an image, contributing to advancements in assistive technology and bridging the communication gap between signers and non-signers. The model was deployed on a web-based interface for real-time usability, making sign language recognition more accessible and practical for everyday interactions. In the experiment, the multi- model was tested using appropriate metrics in which accuracy of 99.90%, AUC of 99.99%, loss of 0.03%, recall of 99.90%, precision of 99.90% were obtained and independent test set that was used on the model showed impressive results. It was observed from this study that the developed approach performed excellently well in predicting sign language alphabets in an image. We recommend the applicability of this work in different sector for effective communication.