Performance And Analysis Of Handwritten Tamil Character Recognition Using Artificial Neural Networks

Research Article
Rajasekar M., Celine Kavida A and Anto Bennet M
DOI: 
xxx-xxxxx-xxxx
Subject: 
science
KeyWords: 
Optical Character Recognition (OCR), Computer handwritten character recognition (HCR), Artificial Neural Networks (ANN),
Abstract: 

Computer handwritten character recognition (HCR) system can improve the human computer interaction and better integrate computers into human society. HCR and optical character recognition (OCR) in a more general context are an integral part of pattern recognition. At the early stages of research and development of pattern recognition most of the researcher investigated the subject of OCR. One of the main reason was because characters were very handy to deal with, since most of the time characters are defined in a two dimensional lattice which have two states, so it was commonly thought that this problem could be easily solved. However, against what was the expectation after some initial progress, great difficulty in solving this problem surfaced. And even today with large scale computer power available and high-quality scanners, OCR still poses some interesting and difficult problem to be solve definitely. Another point of interest is that optical character recognition is rather a universal problem in that it includes essential problems of pattern recognition which are common to all other topics. Thus making it an interesting problem to analyze in views of understanding other more complex problems in pattern recognition and analysis. In this proposed work, introduce the problem of handwritten Tamil character recognition including a historical background on the subject. A review of computer handwritten recognition aims and application is studied, followed by description of previous method and techniques. Handwritten Tamil character recognition, two different approaches in trying to deal with this problem is studied. A moment based feature extraction technique and a coding scheme based on the neighborhood relation are developed. The fuzzy c-means clustering algorithm is used as the method for data reduction. And Artificial Neural Networks (ANN) are applied for the recognition process.