Health Informatics

Permanent URI for this collection

http://etd.aau.edu.et/handle/123456789/204

Browse

Browsing Health Informatics by Subject "Amharic Braille Documents"

Now showing 1 - 1 of 1
  • No Thumbnail Available
    Item
    Recognition of Amharic Braille Documents
    (Addis Ababa University, 2010-06) Chekol, Ebrahim; Assabie, Yaregal (PhD)
    There are more than half million visually impaired people in Ethiopia consisting of students, teachers, artists, lawyers, and also employees who have significant contribution in politics, religious, economics and social affairs of the society. For these people Braille is the means for codifying their knowledge. However, most of their work still remained in their Braille and of course accessed only by those who read and/or write Braille. Since 1924, there are a significant number of old Braille documents produced and used by visually impaired society throughout the country. From this, very insignificant number of Braille has actually reached to the vision society creating communication gap between the vision and visually impaired people. In addition to these, at present, blind students attend regular schools together with their sighted peers in the elementary, secondary and tertiary levels. However, it is difficult, among others, to take quizzes and examinations without the help of others (vision students) in reading and writing, and also in submitting assignments for their teachers. In general, there is no means of written communication between blind and sighted people. In this study an attempt is made to explore the possibility of developing an OBR (optical Braille recognition) system for real life single sided Amharic Braille documents. The system is implemented using artificial neural network for 238 Amharic characters, 10 Arabic numerals and 19 punctuation marks. Gaussian filtering with adaptive histogram equalization and morphological operations are used so as to detect and remove the noises in the real life Amharic Braille documents. The noise detection and removal, and thresholding algorithms are integrated with the previous algorithm implemented for recognition of clean Amharic Braille documents. The present system achieves an accuracy of 95.5%, 95.5%, 90.5%, and 65% for clean, small level noisy, medium level noisy and high level noisy Braille documents respectively. This research approach to control noise in the real life Braille document to a great extent, however, there is some addition and deletion of dots.