A Generalized Approach to Amharic Braille Recognition

Wendmu, Amanuel (2015) A Generalized Approach to Amharic Braille Recognition. Masters thesis, Addis Ababa University.

PDF (A Generalized Approach to Amharic Braille Recognition) Amanuel, worndimu_2015.pdf - Accepted Version Restricted to Repository staff only Download (2MB) | Request a copy

Abstract

In 1825 Louis Braille invented the writing system braille which is the preferred means of written communication for blind people around the world. In Ethiopia, there has been a large production of Amharic braille documents from the time of its introduction in 1934. Since Braille documents are usually read and understood only by visually impaired and those others who have learned braille only, their literatures are highly restricted from reaching the sighted society. Optical Braille Recognition (OBR) is an important technology to bridges the communication gap that exists between the blind and the sighted people. In addition, OBR serves as a technology to preserve documents that only exist in Braille form and can also serve as a simple mechanism for braille copy. Past researches carried out locally have made an attempt to develop an ABR system for Amharic braille recognition. There were also some efforts made to improve the performance of the Amharic ABR system. As a continuation to those previous efforts, this research presented an approach for recognition of double-sided Amharic braille documents. In this regard, image thresholding and segmentation techniques are explored and adopted in this work with some improvements. An adaptive thresholding algorithm that uses the pixel values on every image is used to compute two threshold values that segment the Braille image into bright, dark and background. A moving window scanning technique that scan every column of an image and extract the dot regions for the front and back side of the image and put them into two separate arrays is also adopted and improved. In addition, an improved grid construction technique is designed that automatically determines the average dot area and exclude dots which are smaller or bigger in size that can affect the performance of the grid construction. The grid construction process has incorporated techniques for refining itself when any error exists. Braille cells are formulated by using the location of six intersection points in the constructed grid. Finally, a lookup table is developed for every character in Amharic and features of braille cells are matched against corresponding Amharic print characters. The system is tested using database from scanned clean, average noisy and high level noise Amharic braille documents. The developed system works with 100% accuracy for both single-sided and double-sided clean braille documents. However, as the level of noise increases the performance of that system fails. Overall, an accuracy level of 99.27% for single-sided and 88.16% for double-sided braille documents is achieved. Degradation of braille document and skewness has been the major challenges. The results achieved show that the work is promising. However, to further improve the system, areas that need further consideration in future such as improved image filtering, skew angle correction, format preservation and post recognition correction are recommended.

Item Type:	Thesis (Masters)
Subjects:	P Language and Literature > P Philology. Linguistics P Language and Literature > PL Languages and literatures of Eastern Asia, Africa, Oceania Q Science > QA Mathematics > QA75 Electronic computers. Computer science Q Science > QA Mathematics > QA76 Computer software
Divisions:	Africana
Depositing User:	Selom Ghislain
Date Deposited:	19 Jun 2018 14:23
Last Modified:	19 Jun 2018 14:23
URI:	http://thesisbank.jhia.ac.ke/id/eprint/4705

Actions (login required)

View Item