OCR systems based on neural network

Abstract

This paper deals with the recognition of handwritten text. It is mainly discussing improving nowadays OCR systems. In detail is this article focused on the possibilities of implementing the neocognitron network in this improvement. Next part deals with the problems of document processing, recognition of individual characters and subsequent search for whole words against the dictionary. Main goal of this work is to invent new principles in the field of processing hand written text especially focused on text with language specifics like diacritics.