Person with vision is able to read any document even knowing title, subsection, and paragraph as the ways. In the case of a visually impaired person, it was possible only for some other person giving reading assistants. There must be a provision for visually impaired persons who are supposed to get assistance to read content of documents or books with their title, subtitle, paragraph etc. as per their wish and themselves using some Machine learning (ML) or Deep Learning (DL) algorithms. Some work has been proposed to assist visually impaired persons. With the help of these works are able to scan documents using Object Character Recognition (OCR) technique and with the help of voice assistants, it can be listened to by them. The limitation is that the OCR technique does not identify and categorize as title, sub title and paragraph etc. The idea of this proposed work was that it is essential to identify and categorize reading image based content or document into title, sub title and paragraph within a single image. This solution would be very useful to visually impaired people to read such image content as per their choice and that they can get only what content they want to read and which title, which subtitle and paragraph and all. In this proposed work, this proposed work was coined the Enhanced OCR (EOCR) method. The experiment was conducted and accuracy of categories image content into title, subtitle, paragraph etc. for 10 in count. The OCR method’s accuracy for the same work was measured 10 in count. Later for precision, sensitivity was also noted for both proposed and existing algorithms. With the aid of SPSS tool, performance comparison was carried out with independent samples t test and graph builder modules. The mean accuracy, mean precision and mean sensitivity were measured using proposed EOCR as 97.899%, 97.404% and 97.933% respectively and for OCR algorithm 94.059%, 94.925% and 95.387% respectively. The proposed EOCR method has gained more novelty to recognize images into title, subtitle and paragraph content than the OCR algorithm parentheses.

Deep learning; machine learning; feature extraction; enhanced optical character recognition; visually impaired person assistance