Recognition of multifont Farsi/Arabic characters using a fuzzy neural network

In this paper an algorithm is developed for recognition of printed Farsi characters with various fonts, irrespective of size, rotation and stork. The system uses pseudo-Zernike moments as input features and the classifier consists of a complex of neural networks (NN) and fuzzy neural networks (FNN). The advantage of using FNN is it's ability to classify similar patterns. The performance of the system is evaluated on a database consisting of more than 3700 character samples. The achieved accuracy is 99.85%.