Automatic Computer Vision System to Help the Blind or Visually Impaired by Converting the Written Arabic Text to Audible Style

The goal of this research is to design and implement a portable computerized system that helps the blind or visually impaired to hear Arabic texts which are printed or written by free hand. The system captures an image of Arabic text as an input. The complexity of Arabic language makes the research in recognition of printed and handwritten Arabic text to be untouched for many years compared with other languages such as English and Chinese. In the last years a significant number of researches has been carried out to recognizing printed and handwritten. However, Arabic characters recognition is still an open research field due to its cursive nature. 
The proposed system first aligns the taken image to a prototype template. Then, it subtracts the background. Foreground image will be segmented into lines. Each line is segmented into isolated words. The word will be segmented to sub-words. The word or sub-word will be segmented into isolated characters. Finally, the features of each segmented character are calculated in features vector. This vector will be compared with a pre-defined templates of feature vectors for all of the Arabic letters. The last step is to pass the segmented word to a program that synthesized and pronounce this word. The proposed framework has been evaluated using over dozen thousands of different Arabic words. The experimental results show that average recognition efficiency of the proposed framework is highly significant compared with other methods.