Handwritten character recognition is a subject of importance in these days. Artificial Neural Networks (ANNs) are very much in demand in order to accomplish the task and that is why mass research is also going on in this field. This paper is an approach to identify handwritten characters by observing the gradient of the pixel densities at different segments of the handwritten characters. Different segments of the characters are observed carefully with the help of generated computer programs and rigorous experiments. It is found that the pixel densities at various segments of the character image matrix of different alphabets vary. The gradient of the pixel densities in these segments are used to form unique codes for different alphabets, which are found standard for different variations of same alphabet. Generation of unique codes actually extracts out common features of a particular alphabet written by one or more individuals at different instants of time. The unique codes formed for different alphabets are used to recognize different test alphabets. The method developed in this paper is a feature extraction technique which uses self organizing neural network, where supervised learning is not required.

Artificial Neural Networks; Pixel Density Gradient; Segments; Handwritten Character

[1] G Robinson, �The multiscale technique�, Available: http://www.netlib.org/utk/lsi/pcwLSI/text/node123.html, Page No (1), Mar 1995.
[2] TCTS Website, �Handwritten character recognition�, Available: http://tcts.fpms.ac.be/rdf/hcrinuk.htm, Page No (1), Accessed: 2010 (June).
[3] V Ganapathy and K L Liew, �Handwritten character recognition using multiscale neural network training technique�, Proceedings of World Academy of Science, Engineering and Technology, Page No (29-37), May 2008.
[4] S Arora, D Bhattacharjee, M Nasipuri, D K Basu and M Kundu, �Combining multiple feature extraction techniques for handwritten Devnagri character recognition�, IEEE Region 10 Colloquium and the Third ICIIS, Page No (1-6), Dec 2008.
[5] D Singh, S K Singh and M Dutta, �Handwritten character recognition using twelve directional feature input and neural network�, International Journal of Computer Applications (0975 � 8887), Page No (82-85), 2010.
[6] R K Mandal and N R Manna, �Handwritten English character recognition using row-wise segmentation technique (RST)�, International Symposium on Devices MEMS Intelligent Systems Communication (ISDMISC), Proceedings published by International Journal of Computer Applications� (IJCA), Page No (5-9), April 2011.
[7] R K Mandal and N R Manna, �Handwritten English character recognition using column-wise segmentation of image matrix (CSIM)�, WSEAS Transactions on Computers, E-ISSN: 2224-2872, Volume 11, Issue 05, Page No (148-158), May 2012.
[8] G N Swamy, G Vijay Kumar, �Neural networks�, Scitech, India Pvt Ltd, ISBN: 8183710557|9788183710558, 2007.
[9] L Fausett, �Fundamentals of neural networks, Architectures, Algorithms and Applications�, Pearson Education, Fourth Edition, ISBN: 978813170053-2, Page No (19-114), 2009.
[10] A Roy and N R Manna, �Character recognition using competitive neural network with multi-scale training�, UGC Sponsored National Symposium on Emerging Trends in Computer Science (ETCS 2012), Page No (17-20), Jan 2012.
[11] A Roy and N R Manna, �Competitive neural network as applied for character recognition �, International Journal of advanced research in Computer Science and Software Engineering, Volume 02, Issue 03, Page No (06-10), 2012.
[12] D S Rajput, R S Thakur and G S Thakur, �Clustering approach based on efficient coverage with minimum weight for document data�, IJCSE, International Journal of Computer Sciences and Engineering, Volume 01, Issue 01, Page No (06-13), 2013.