Hand Written Character Recognition & Comparisons Using VPP, HPP, ZDCT & CCH Extraction Techniques

Character recognition covers all types of machine simulation of human reading which forms a part in various application domains. The aim of character recognition is to translate human readable characters to machine readable characters. In this work, an automatic character extraction and recognition system has been developed for extracting the character level information from handwritten application forms and then recognizing them. Proposed automatic character extraction system can provide more practical handwritten character database in an easy, time efficient and automated way and also gives an approach for automatic handwritten document processing considering passport, insurance and banking applications. It utilizes the vertical and horizontal projection profiles (VPP-HPP) and includes binarization, field extraction, skew correction, character extraction and size normalization at various levels for isolated character extraction. An in-house database of isolated handwritten numerals and characters is successfully collected by using this automatic approach. To recognize the extracted characters, isolated numeral and character recognition systems have been built by using VPP-HPP, zonal discrete cosine transform (ZDCT) and chain code histogram (CCH) features. The analysis of experimental results indicate that small amount of misclassification is due to the large shape similarity across the confusing pairs of numerals and characters.