Recognition of Alphabet Characters and Arabic Numerals Using Back Propagation Neural Network

摘要

This paper proposed an optical recognitionfor Arabic numeric and Alphabet Characters, anattempt was made to recognize the equivalent Englishnumeric numbers from the Arabic dataset. Alphabetcharacters were tested with the intention of recognizingevery of its main subject character when a probe testcharacter is supplied. The problem in characterrecognition majorly is in the variation and distortion ofhandwritten characters. Since different people may usediverse poses and methods of handwritten, anddirection to draw the same shape of the characters.Sixty three datasets with nine different Arabicrepresentations by 7 people were used for trainingthe Neural Network and 30 different handwrittenArabic numeral characters were used for testing.While the alphabet characters took 1650 for trainingand 30 different alphabets was used to test the system.The recognition system performed effectively for thetwo characters, which gave better recognitionaccuracy of 91.66% for Arabic numerals and 92% forAlphabet charactesr respectively.