Braille to text conversion for Afaan Oromo Using Deep Learning

Abstract

In the year in 1825, Louis Braille created the Braille writing system, the preferred written communication system for the visually impaired around the world. Since its introduction in Ethiopia in 1986, a large number of Afaan Oromo Braille documents have been produced. In this study, an attempt was made to explore the possibility of creating OBR (Optical Braille recognition) system for real-life single-sided Afaan Oromo Braille documents. Assistive Braille technology has existed for many years intending to help visually impaired people perform common tasks such as reading, writing, and communicating with others. Such technologies aim to help the visually impaired better adapt to the visual world. However, there is a clear gap in current technology for balanced two-way communication between visually impaired and sighted people, as little technology allows non-visually impaired people to understand the Braille system. This study presents the paper in this language using convolutional neural network (CNN) model and Proportional Character Segmentation Algorithm (PCSA) to convert Braille images into Afaan Oromo text. In addition, a new dataset containing of 23,000 labeled Braille images consists of 37 Braille characters, including letters, punctuation and numbers corresponding to 71 different Afaan Oromo class. The performance of the convolutional neural network model gave a prediction accuracy of 96% on the test set. The functionality performance of this artificial intelligence (AI) based recognition system can be tested in the future with an accessible user interface.