Design of CNN architecture for Hindi Characters

Madhuri YADAV, Ravindra KR PURWAR, Anchal JAIN

Abstract


Handwritten character recognition is a challenging problem which received attention because of its potential benefits in real-life applications. It automates manual paper work, thus saving both time and money, but due to low recognition accuracy it is not yet practically possible. This work achieves higher recognition rates for handwritten isolated characters using Deep learning based Convolutional neural network (CNN). The architecture of these networks is complex and plays important role in success of character recognizer, thus this work experiments on different CNN architectures, investigates different optimization algorithms and trainable parameters. The experiments are conducted on two different types of grayscale datasets to make this work more generic and robust. One of the CNN architecture in combination with adadelta optimization achieved a recognition rate of 97.95%. The experimental results demonstrate that CNN based end-to-end learning achieves recognition rates much better than the traditional techniques.

Keywords


hindi character recognition, handwritten character recognition, deep learning, cnn, optimization algorithms

Full Text:

PDF

References


Acharya, S., Pant, A. K., and Gyawali, P. K., 2015. Deep learning based large scale handwritten Devanagari character recognition. In 2015 9th International Conference on Software, Knowledge, Information Management and Applications (SKIMA), pages 1-6. doi:10.1109/SKIMA.2015.7400041.

Belhe, S., Paulzagade, C., Deshmukh, A., Jetley, S., and Mehrotra, K., 2012. Hindi Handwritten Word Recognition Using HMM and Symbol Tree. In Proceeding of the Workshop on Document Analysis and Recognition, pages 9-14. ACM. ISBN 978-1-4503-1797-9. doi:10.1145/2432553.2432556.

Bottou, L., 2012. Stochastic Gradient Descent Tricks. In Neural Networks: Tricks of the Trade, volume 7700, pages 421-436. ISBN 9783642352898. doi:10.1007/978-3-642-35289-8.

Deepti Khanduja, S. P., Neeta Nain, 2015. Hybrid Feature Extraction Algorithm for Devanagari Script. ACM Transactions on Asian and Low-Resource Language Information Processing (TALLIP), 15:2:1-2:10.

Glorot, X. and Bengio, Y., 2010. Understanding the difficulty of training deep feedforward neural networks. In Proceedings of the Thirteenth International Conference on Artificial Intelligence and Statistics, volume 9, pages 249-256. PMLR.

Gyanendra K.Verma, P. K., Shitala Prasad, 2011. Handwritten Hindi Character Recognition Using Curvelet Transform. In Information Systems for Indian Languages, pages 224-227. Springer.

Hanmandlu, M., Grover, J., Madasu, V. K., and Vasikarla, S., 2007. Input Fuzzy Modeling for the Recognition of Handwritten Hindi Numerals. In Information Technology, 2007. ITNG ’07. Fourth International Conference on, pages 208-213. IEEE.

Haykin, S., 1998. In Neural Networks: A Comprehensive Foundation, 2. Prentice Hall.

Kekre, H. B., Thepade, S. D., Sanas, S. P., and Shinde, S., 2013. Devnagari Handwritten Character Recognition using LBG vector quantization with gradient masks. In 2013 International Conference on Advances in Technology and Engineering (ICATE), pages 1-4. doi:10.1109/ICAdTE.2013.6524768.

Kingma, D. and Jimmy, B., 2014. Adam: A method for stochastic optimization. In International Conference on Learning Representations, pages 1-15.

Lecun, Y., Bottou, L., Bengio, Y., and Haffner, P., 1998. Gradient-based learning applied to document recognition.

Proceedings of the IEEE, 86(11): 2278-2324. ISSN 0018-9219. doi:10.1109/5.726791.

Madhuri Yadav, R. K. P., 2018. Hindi handwritten character recognition using oriented gradients and Hu- geometric moments. Journal of Electronic Imaging, 27(5):051216.1-051216.11.

Prasad, S., Verma, G., Singh, B., and Kumar, P., 2012. Basic handwritten character recognition from multi-lingual image dataset using multi-resolution and multi-directional transform. 10.

Sarkhel, R., Das, N., Das, A., Kundu, M., and Nasipuri, M., 2017. A multi-scale deep quad tree based feature extraction method for the recognition of isolated handwritten characters of popular indic scripts. Pattern Recognition, 71: 78-93. ISSN 0031-3203. doi:https://doi.org/10.1016/j.patcog.2017.05.022.

Srivastava, N., Hinton, G., Krizhevsky, A., Sutskever, I., and Salakhutdinov, R., 2014. Dropout: A Simple Way to Prevent Neural Networks from Overfitting. J. Mach. Learn. Res., 15(1): 1929-1958. ISSN 1532-4435.

Tieleman, G. H., 2012. Lecture 6.5-rmsprop: divide the gradient by a running average of its recent magnitude. In COURSERA: Neural Networks for Machine Learning.

Yadav, M. and Purwar, R., 2017. Hindi handwritten character recognition using multiple classifiers. In 2017 7th International Conference on Cloud Computing, Data Science Engineering - Confluence, pages 149-154. doi:10.1109/CONFLUENCE.2017.7943140.

Zeiler, M. D., 2012. ADADELTA: an adaptive learning rate method. In arXiv preprint arXiv:1212.5701.




DOI: http://dx.doi.org/10.14201/ADCAIJ2018734762





Creative Commons License
This work is licensed under a Creative Commons Attribution 3.0 License.

Clarivate Analytics