Conversion of Braille Document to Regular English Document using VGG16 BC-CNN Model

Braille metallic documents are the only medium the visually impaired people will use for reading. Due to their heavy bulky physical nature, it is cumbersome to move them manually from one area to other over the globe. Apart from distribution, it is also very tedious to maintain them as it is a natural tendency of plates to get wear and tear in different weather conditions. Hence by translating the Braille documents to natural text, will help in distributing it over the network across globe as well as it can be maintained in digital form for later reproduction into Braille form. This research paper presents the proposed work that translate a Braille document to the corresponding English sentence. The preprocessing stage includes the removal of unwanted dots and skew detection and removal methods. Segmentation stage performs the row and column segmentation to locate the cell pattern in Braille image, the mapped cells are cropped and stored in Image array. Feature extraction stage uses the cropped Braille cell image array as input to Braille classifier model built using VGG16 CNN model. The model maps these images it to the corresponding characters and thus generate the English statement.