Improved Camera-Based Text Reading Assistant System Using Digital Image Processing Techniques

Abstract

This paper proposes an Intelligent Text Reading Assistant (ITRA) that works to read the text in any image to serve visually impaired people and help them to read their surroundings. The proposed system is designed to be built on a Raspberry Pi connected with a camera that is used to capture the input image. The input image is enhanced by applying image processing techniques. The Tesseract optical character recognition (OCR) engine embedded in the Raspberry Pi searches for the text in an improved image and converts it into digital text file. The text file is then converted to mp3 file using Google Text to Speech (gTTS) technology. The results obtained through system implementation were analyzed by following the categorizing approach that works to interpret results by grouping found data into categories where the overall accuracy of the system was 97% in English and 85% in Arabic. The conducted analysis showed that the system performs well with English text but has a moderate level of accuracy with Arabic text. It also showed that the speech produced for English text is as clear as natural human voice whereas for Arabic audio, it is like a machine sound.