Real Time Implementation of Optical Character Recognition System for Multi line Refreshable Braille Display

Abstract

Braille way of learning was invented to empower visually impaired people and make them self-reliant. Most of the them remain illiterate because of their limited access to Braille material and Braille related learning devices. Socio-economic condition of blind individuals is the most common reason that deprive them from getting access to education, Assistive Technologies, and related aids. Their interaction with the surroundings, communication with their peers, and independent lifestyle gets severely hampered. newlineThis thesis proposes a unique hardware design for a Refreshable multi-line Braille reading device with Optical Character Recognition (OCR) capabilities for learning and reading text in Braille codes. The device uses a microprocessor board to seamlessly integrate stepper/servo motors with OCR algorithms. A thin flexible sheet coated with rubber is used as a display surface on which the raised point for Braille codes is repeatedly formed and deformed. The device is designed in such a way that the material used for its construction are low cost which makes them economical and affordable. The device was evaluated in the lab setup and showed promising results, and had prospects of becoming a vital Assistive Technology for vision impaired people. newlineThis work also proposes text detection in natural scene images by the Intelligent Grouping Method and Text shape structure analysis algorithm. It is followed by character recognition using novel TexNet, a state-of-art architecture based on Convolutional Neural Network (CNN) which performed with 93.27% accuracy rate. The OCR system was originally designed to extract text from real-time natural scene images that are captured through a portable camera attached to the device. The extracted text is finally converted to Braille codes and displayed on a tactile display surface.