الفهرس 
IntroductionOnly 14 pages are availabe for public view
 |  | from | 119 |  |  |
| | | from | 119 | | |
Abstract
Visually Impaired (VI) people suffer from many difficulties when accessing printed text using existing technologies. These problems may include text alignment, focus, accuracy, software processing speed, mobility, and efficiency. The issues become more complicated when accessing text documents in various situations. Current technologies such as flatbed scanners and OCR programs need to scan an entire page before processing text. It is not feasible to help VI people read the text in real-time using these previously mentioned technologies. Recently, VI people prefer mobile devices because of their handiness and accessibility, but they have problems focusing on the printed text’s using a mobile camera. This thesis introduces a real-time Arabic text-reading algorithm for VI or blind People. The proposed algorithm uses a wearable device for hand fingers. It is designed as a wearable ring attached to a tiny webcam device. The VI person wears the ring on his finger. The attached camera captures the printed Arabic text and passes it to the Arabic OCR algorithm.
Much research in OCR for various languages was proposed, but Arabic OCR is still evolving because of the Arabic word’s syntax and structure’s complex nature. An efficient character recognition technique is constructed for variant fonts of the Arabic text. Arabic OCR has six main stages: preprocessing, line segmentation,
Abstract
v
word segmentation, character segmentation, feature extraction, and Classification. We concentrate on choosing an algorithm that achieves high accuracy and low time in Arabic OCR in every stage. The accuracy of the proposed OCR is improved as compared to existing Arabic OCR. Experimental results demonstrate the feasibility of the proposed algorithm. Our Algorithm can also recognize English text by using the proposed English OCR. A new feature extraction algorithm based on the character image’s Statistical and Regional Features (SRF) is presented. The proposed approach succeeded in obtaining the original data’s essential information and representing it in a lower dimensionality space. The proposed algorithm gives a feature vector with a length of 85 as its output. The time required for extracting the statistical features is very high. Therefore, the Universe of discourse is used to speed up retrieval. Feature selection is performed using the Euclidean distance. from the practical analysis, the new feature extraction algorithm’s accuracy is improved compared to existing algorithms.
Finally, the recognized characters are translated into speech using the Text-To-Speech (TTS) technology. Experimental results demonstrate the feasibility of the proposed algorithm. It achieved an accuracy of 96 % for Arabic character recognition and 98.5% for English character recognition.