Abstract

In this paper, Braille character recognition (BCR) module is proposed that translates a single side Braille documents into English text and voice signal to help blind peoples saving their documents and hear it whenever they need the module is consisting mainly of two stages: the preprocessing stage and the recognition stage. In the 1st stage different thresholds and mask sizes were tested, and then the preprocessed image will be used in the recognition stage. Modify Multi-Connect Architecture (MMCA) and Modify Bidirectional Associative Memory (MBAM) algorithms were used to get the English document. A comparison is made between the two algorithms to get the best results. The implemented (MMCA) algorithm achieved average accuracy for correct letters was 98.26%, average accuracy for correct words was 95.11% and average processing time around 11.5 sec per page, The implemented (MBAM) algorithm achieved average accuracy for correct letters was 91.87%, average accuracy for correct words was 51.26% and average processing time around 3.4 sec per page.

Authors and Affiliations