Automatic Processing of Handwritten Arabic Forms Using Neural Networks

Abstract — Among the vast range of off–line optical character recognition applications is the machine processing of forms. The objective is to imitate the human ability to read but at higher speed. This paper presents a neural network based system to recognise handwritten Arabic characters collected from more than 500 forms. Wavelet coefficients extracted from the character samples are used to tune the neural network weights. The performance of the system is assessed using a corpus that includes 15800 character samples.