Printed and Handwritten Arabic Characters Recognition and Convert It to Editable Text Using K-NN and Fuzzy Logic Classifiers


In this paper we suggest an off-line isolated Arabic characters recognition system and this system has an ability to recognize printed and handwritten Arabic character and then convert these characters to printed text by mixed image processing techniques and artificial intelligent system. These techniques used to find rigid features for each Arabic character to distinguish it from another characters in Arabic language. K-Nearest Neighbors (K-NN) classifier was used to classify the printed character and fuzzy logic to classify handwritten Arabic character. Different font of printed character have font type (Arabic transparent, Times New Roman, Arial, simplified Arabic fixed) and font size 14 in order to test the quality of our system, each printed character in Arabic alphabet entered nine times, (6) of them in Arabic transparent font while other (3) is in Times New Roman, Arial, simplified Arabic fixed respectively while each handwritten character enter six times three of them used to training and the remaining (3) are used to testing. So 324 printed character are entered to our system ,the system successes in recognize 301 character form printed characters with recognition ratio is 92.9%.while 216 handwritten Arabic character entered to our system which successes in recognize 208 character of them with recognition ratio is 96.6%. Elapsed time in execute our system to perform recognition process for one character is 0.04 seconds.