Development of OCR system on android platforms to aid reading with a refreshable braille display in real time

Individuals with visual impairment are limited in terms of communication, interaction and personal autonomy due to the lack of literature in Braille which is mainly attributable to economic reasons. This paper proposes a reading system for visually impaired persons using a portable device. This work proposes and evaluates a combination of segmentation, feature extraction and machine learning techniques to achieve the best conversion of text to braille as quickly and accurately as possible. The experiments showed that the Central Moments extractor with Multi Layer Perceptron were the best combination for the OCR system developed with 99.86% accuracy and 99.93% specificity. Furthermore, we assess the portable device usability with elementary teachers and with teachers and students in an association of the blind. The results of this system can contribute to improved socialization between visually impaired persons and stimulate their intellectual health.