A Novel and Efficient Algorithm to Recognize Any Universally Accepted Braille Characters: A Case with Kannada Language

Abstract

A Braille document image is a collection of dots. The position of the dot and relative-ness of the dot with other dots gives different Braille characters. It is challenging, to separate the character lines, words and characters from a Braille document. This paper presents an Optical Braille character recognition system for both machine punched and hand punched Kannada Braille text documents [21]. Standard spacing between the characters and lines are used to segregate the dots. Dot mesh is created and character box is identified. Once character box is identified an efficient look up method is designed to identify the equivalent normal Kannada character. A unique value for the Braille character is generated and the Braille character is matched to the corresponding normal Kannada character in one shot. A Braille character is made of 6 dots combination and hence only 26=64 different combinations are possible. Recognized character is classified into one of the 64 possible classes. Identifying the dot position inside a character box is done using the dot mesh and by computing the centre position of all the objects inside the character box.