An Easy Approach of Integrating HTR Models into Hand Written Character Recognition
Abstract
Automatic handwritten
character recognition is a subject that receives much attention at present. One
of the main drivers behind current research is the capacity to quickly interpret
tiny handwriting samples such as those seen in checks and envelopes. Many
(offline) hand-written Text Recognition (HTR) systems research have been
conducted to create state-of-the-art models for small enterprise line
recognition. But it presents significant problems to add HTR capabilities to a
multiple OCR system. This article deals with three issues related to systems
such as data, efficiency and integration. The Project is a computer based
programme that minimises effort in converting the handwritten script
photographs into text documents. The problems are addressed through the use of
online handwriting data for the online recognition system for a large-scale
manufacturing. We present our pipeline of picture data creation and investigate
how HTR models may be built using online data. We show that the data
considerably enhance models in the circumstance that just a few numbers of
actual pictures, generally the case with HTR models, are available. It allows
us to considerably decrease the costs of a new script. Secondly, we present a
model for line recognition based on non-recurring connectivity neural networks.
We are exploring this approach in order to develop an excellent English written
word recognition system based on the recognition of character. Lexicon
post-processing is used to increase the overall accuracy of recognition. There
are several approaches accessible for the extraction and training, each with
its own superiorities and limitations, of CR systems in literature. With the
LSTM models, the model achieves equivalent precision while enhancing
parallelism in training and inferences. Finally, we are offering an easy
approach of integrating HTR models into OCR. This is a solution for bringing
HTR into a wide-ranging OCR.
Country : India
1 Manju Padidela
Associate Professor, Department of Electronics and Communication Engineering, Malla Reddy College of Engineering for Women, Hyderabad -500100, Telangana, India
Cho, Sung-Bae. “Pattern recognition
with neural networks combined by genetic algorithm.”Fuzzy Sets and Systems,
Vol. 103(2), April 1997. pp. 339-347.
Gader, Paul D., et al. “Neural and
Fuzzy Methods in Handwriting Recognition.” Computer, February 1997. pp. 79-86.
Liu, Cheng-Lin, and Nakagawa, Masaki.
“Handwritten Numeral Recognition Using Neural Networks: Improving the Accuracy
by Discriminative Training.” Proceedings of the Fifth International Conference
on Document Analysis and Recognition, September 1999. pp. 257-260.
Luger, George F., and Stubblefield,
William A. Artificial Intelligence: Structures and Strategies for Complex
Problem Solving, (2nd Edition). Benjamin/Cummings Publishing Company, Inc.,
California, 1993, pp. 516-527.
Molen, A.E., and Veelenturf, L.P.J.
“On-line signature validation using Kohonen’s Neural Network.” Neural Networks:
Best Practice in Europe. World Scientific Publishing Co. Pte. Ltd., Singapore.
pp. 145-148.
Russell, Stuart J., and Norvig,
Peter. Artificial Intelligence: A Modern Approach. Prentice-Hall, Inc, New
Jersey, 1995. pp. 567- 587.
Xiao, Xu-Hong, and Leedham, Gary.
“Signature Verification by Neural Networks with Selective Attention.”Applied
Intelligence, Vol. 11(2), Sept/Oct. 1999. pp. 213-223.
Stephane Armand, Michael Blumenstein
and Vallipuram Muthuk kumarasamy, “Off-line Signature Verification based on the
Modified Direction Feature”, 2006 International Joint Conference on Neural
Networks, pages 684-691.
