Two Methods for Recognition of Hand Written Farsi Characters

摘要: Optical character recognition (OCR) is one of the active bases sample detection topics. The current study focuses on automatic and hand written Farsi characters. For this purpose; we proposed two different methods based neural networks a special post processing approach to improve rate uppercase letters. In first method, extracted wavelet features from borders images learned network these patterns. second divided input characters into five groups according number their components used set appropriate moment in each group classified by Bayesian rule. post-processing stage, some structural statistical were employed decision tree classifier reduce misrecognition rate. Our experimental results show suitable for both methods.