Real-time document reader assistance app for the visually impaired

Choong, Darren Yu Xuen (2025) Real-time document reader assistance app for the visually impaired. Final Year Project, UTAR.

Preview

PDF Download (5Mb) | Preview

Abstract

Due to the fast-paced lifestyle today, blind people find it hard to access documents independently and efficiently. The growing demand for accessible technology has brought attention to the challenges faced by visually impaired individuals when accessing and reading physical and digital documents. Furthermore, most assistive technology requires a subscription plan, therefore leading to financial difficulty for visually impaired individuals. To tackle these problems, a real-time document reader assistance application using Flutter is developed for free to use. It will be developed on both the iOS and Android platforms. It will also be focused more on the targeted audience, which is the visually impaired individuals. Hence, the proposed solution will integrate OCR text recognition techniques for text extraction using Google ML Kit. Additionally, a transfer learning on pre-trained model (YOLOv8n) will be used to detect documents in the camera preview. Moreover, Text-to-Speech (TTS) will be implemented for real-time audio feedback. Furthermore, an AI chatbot will be implemented to aid users in further clarification of the recognized text from the document. Speech-to-Text (STT) will also be implemented to convert audio from the user into text when questioning the AI chatbot assistance without typing. This project development process follows the Agile methodology, which involves several phases such as planning, analysis, design, implementation, and testing, with continuous iterations. The result of this project will serve as an innovative tool that improves the accessibility of information for visually impaired users and contributes to the MAHAL application. In conclusion, this project will provide a significant contribution to enhancing the independence and accessibility of visually impaired users.

Actions (login required)