Recent Progress in Arabic Sign Language Recognition: Utilizing Convolutional Neural Networks (CNN)

The advancement of assistive communication technology for the deaf and hard-of-hearing community is an area of significant research interest. In this study, we present a Convolutional Neural Network (CNN) model tailored for the recognition of Arabic Sign Language (ArSL). Our model incorporates a met...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:BIO web of conferences 2024-01, Vol.97, p.73
Hauptverfasser: Hassan, Mosab. A., Ali, Alaa. H., Sabri, Atheer A.
Format: Artikel
Sprache:eng
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:The advancement of assistive communication technology for the deaf and hard-of-hearing community is an area of significant research interest. In this study, we present a Convolutional Neural Network (CNN) model tailored for the recognition of Arabic Sign Language (ArSL). Our model incorporates a meticulous preprocessing pipeline that transforms input images through grayscale conversion, Gaussian blur, histogram equalization, and resizing to standardize input data and enhance feature visibility. Principal Component Analysis (PCA) and Linear Discriminant Analysis (LDA) are employed for feature extraction to retain critical discriminative information while reducing dimensionality. The proposed CNN architecture leverages a blend of one-dimensional convolutional layers, max pooling, Leaky ReLU activation functions, and Long Short-Term Memory (LSTM) layers to efficiently capture both spatial and temporal patterns within the data. Our experiments on two separate datasets—one consisting of images and the other of videos—demonstrate exceptional recognition rates of 99.7% and 99.9%, respectively. These results significantly surpass the performance of existing models referenced in the literature. This paper discusses the methodologies, architectural considerations, and the training approach of the proposed model, alongside a comparative analysis of its performance against previous studies. The research outcomes suggest that our model not only sets a new benchmark in sign language recognition but also offers a promising foundation for the development of real-time, assistive sign language translation tools. The potential applications of such technology could greatly enhance communication accessibility, fostering greater inclusion for individuals who rely on sign language as their primary mode of communication. Future work will aim to expand the model's capabilities to more diverse datasets and investigate its deployment in practical, everyday scenarios to bridge the communication gap for the deaf and hard of hearing community.
ISSN:2117-4458
2117-4458
DOI:10.1051/bioconf/20249700073