TRANSLATING ARABIC SIGN LANGUAGE (ARSL) TO TEXT USING ARTIFICIAL NEURAL NETWORKS
Published In: 5TH INTERNATIONAL CONFERENCE ON ADVANCES IN COMPUTING, COMMUNICATION AND INFORMATION TECHNOLOGY
Author(s): LINA ELSIDDIG ABDELRAHIM ELSIDDIG , MAYADA MOHAMED ISMAIL MOHAMED
Abstract: A communication gap exists between the hearing and hearing - impaired communities due to a lack of familiarity wi th the means of communication of each. This research attempts to bridge this distance by creating Arabic Sign Language (ArSL) datasets, which there is a lack of, image processing , selecting a feature extraction method and designing a machine learning class ification system capable of translating Arabic Sign Language (ArSL) to text. The system was implemented on MATLAB 2014a using an Artificial Neural Network that was trained on the morphological features of 100 samples to classify input images into 3 alphabe t classes that achieved an accuracy of 73.3%.
- Publication Date: 03-Sep-2017
- DOI: 10.15224/978-1-63248-131-3-17
- Views: 0
- Downloads: 0
A METHOD FOR EXTRACTING AFFINE INVARIANT LOCAL FEATURES USING A METRIC TENSOR
Published In: 5TH INTERNATIONAL CONFERENCE ON ADVANCES IN COMPUTING, COMMUNICATION AND INFORMATION TECHNOLOGY
Author(s): TAKASHI TORIU
Abstract: This paper proposes a method for extracting local features which is invariant under affine transformations. This method is based on the metric tensor . Using the met ric tensor, an affine invariant smoothing filter and a set of affine invariant differential operator s are construc ted. Then, combining them a set of affine invariant feature extractions is constructed. Effectiveness of the proposed method is confirmed by experiments.
- Publication Date: 03-Sep-2017
- DOI: 10.15224/978-1-63248-131-3-18
- Views: 0
- Downloads: 0