Diacritics Recognition Based Urdu Nastalique OCR System

Improvements and new developments in the field of Artificial Intelligence have opened new horizons in the advancement of machines that originally have limited intelligence. As compared to human brain, machines have already better computational speed and storage however there is still much room to improve the capability to acquire and process data and draw conclusions from it on its own. Optical Character Recognition (OCR) deals exclusively with printed designs and hand written text in nature. Plenty of developments have been made in OCR so far in recognition of Latin, Asian, Arabic and Western texts. As far as Urdu is concerned the work is almost non-existent when compared with the languages cited above. One of its main reasons is the use of extremely complex characters of Nastalique style in Urdu. A methodology for the recognition and processing of the diacritics of Nastalique script is presented in this research work. The proposed technique is effective in recognizing cursive texts with invariant font size of 48. A dataset of 6728 main Urdu Nastalique ligatures is used for the testing purposes which shows that this new technique has the capacity to recognize Nastalique ligatures by having an accuracy of 97.40%. The proposed research work also focuses to improve the existing base mark association process of the Urdu OCR system.