Computationally Efficient Holistic Approach for Handwritten Urdu Recognition using LRCN Model.
Keywords:
Handwritten Urdu, Optical Character Recognition, Urdu Nastaliq Handwritten Dataset, Convolutional Neural NetworksAbstract
Handwritten Urdu text is difficult to recognize because it poses several challenges such as writer-dependent variations in producing different ligature shapes, irregular positioning of diacritics, and similarity in shape of some Urdu characters in writing. Moreover, the formulation and labeling of a huge database of handwritten Urdu is also a challenging task. Due to these challenges, the handwritten Urdu OCR remained the least explored to date. The goal of a writer's adaptive handwritten recognition system is to build a model that improves the recognition of a generic recognition system for a specific author. Few researchers propose the handwritten Urdu datasets but only UNHD is publicly available. Although UNHD claims to have 10000 text lines, only 700 of those lines are unique in terms of semantic content. Moreover, the UNHD contains ligatures of length only up to five characters and doesn’t cover the entire Urdu ligature corpus. Also, the handwritten Urdu recognition techniques proposed by most studies focus on segmentation-based (implicit or explicit) approaches and have not obtained a significant recognition rate on benchmark datasets. Hence an enriched database and an exhaustive recognition technique are needed that can recognize any unconstrained handwritten Urdu text. In this research, we have proposed a new robust holistic handwritten Urdu text recognition technique evaluated on our new handwritten Urdu database UHLD (Urdu handwritten Ligature Dataset). In our proposed technique, CNN extracts features from UNHD and UHLD datasets, and these features train the multidimensional LSTM based recurrent neural network for the classification and recognition of handwritten Urdu text. The technique is computationally efficient and achieved a remarkable accuracy recognition rate of 94.2 % for UNHD and 96.6% for UHLD datasets.
Downloads
References
Naz S, Umar AI, Shirazi SH, Khan SA, Ahmed I, Khan AA. Challenges of Urdu named entity recognition: a scarce resourced language. Research Journal of Applied Sciences, Engineering and Technology. 2014 Sep 15;8(10):1272-8.
Daud A, Khan W, Che D. Urdu language processing: a survey. Artificial Intelligence Review. 2017 Mar 1;47(3):279-311.
Satti DA, Saleem K. Complexities and implementation challenges in offline urdu Nastaliq OCR. In Proceedings of the Conference on Language & Technology 2012 Nov (pp. 85-91).
Khan NH, Adnan A. Urdu optical character recognition systems: Present contributions and future directions. IEEE Access. 2018 Aug 16;6:46019-46.
Dawood, Hassan, Hussain Dawood, and Ping Guo. "Improved Arabic Word Classification using Spatial Pyramid Matching Method."
Wang, L., Chen, K., Long, Y., Mao, X. and Wang, H., 2015, September. A modified efficient certificateless signature scheme without bilinear pairings. In 2015 International Conference on Intelligent Networking and Collaborative Systems (pp. 82-85). IEEE.
Naz, S., Umar, A.I., Ahmed, R., Razzak, M.I., Rashid, S.F. and Shafait, F., 2016. Urdu Nasta’liq text recognition using implicit segmentation based on multi-dimensional long short term memory neural networks. Springer Plus, 5(1), p.2010.
Naz, S. et al., 2015. Segmentation techniques for recognition of Arabic-like scripts: A comprehensive survey. Education and Information Technologies, 21(5), pp.1225–1241. Available at: http://dx.doi.org/10.1007/s10639-015-9377-5.
Din, I.U., Siddiqi, I., Khalid, S. and Azam, T., 2017. Segmentation- free optical character recognition for printed Urdu text. EURASIP Journal on Image and Video Processing, 2017(1), p.62.
Lehal, G.S., 2012, December. Choice of recognizable units for Urdu OCR. In Proceeding of the workshop on document analysis and recognition (pp. 79-85).
Ahmed, S.B., Naz, S., Swati, S., Razzak, I., Umar, A.I. and Khan, A.A., 2017. UCOM Offline Dataset-An Urdu Handwritten Dataset Generation. International Arab Journal of Information Technology (IAJIT), 14(2).
Ahmed, S.B., Naz, S., Swati, S. and Razzak, M.I., 2019. Handwritten Urdu character recognition using one-dimensional BLSTM classifier. Neural Computing and Applications, 31(4), pp.1143-1151.
Husnain, M., Saad Missen, M.M., Mumtaz, S., Jhanidr, M.Z., Coustaty, M., Muzzamil Luqman, M., Ogier, J.M. and Sang Choi, G., 2019. Recognition of Urdu handwritten characters using convolutional neural network. Applied Sciences, 9(13), p.2758.
Hassan, S., Irfan, A., Mirza, A. and Siddiqi, I., 2019, August. Cursive Handwritten Text Recognition using Bi-Directional LSTMs: A Case Study on Urdu Handwriting. In 2019 International Conference on Deep Learning and Machine Learning in Emerging Applications (Deep-ML) (pp. 67-72). IEEE.
Ahmed, S.B., Hameed, I.A., Naz, S., Razzak, M.I. and Yusof, R., 2019. Evaluation of handwritten Urdu text by integration of MNIST dataset learning experience. IEEE access, 7, pp.153566-153578.
Gang Chen and Sargur N. Srihari. 2014. Removing Structural Noise in Handwriting Images using Deep Learning. In Proceedings of the 2014 Indian Conference on Computer Vision Graphics and Image Processing (ICVGIP '14). Association for Computing Machinery, New York, NY, USA, Article 28, 1–8.
DOI: https://doi.org/10.1145/2683483.2683511.
Rehman KU, Khan YD. A Scale and Rotation Invariant Urdu Nastalique Ligature Recognition Using Cascade Forward Backpropagation Neural Network. IEEE Access. 2019 Aug 19; 7:120648-69.
Uddin I, Javed N, Siddiqi IA, Khalid S, Khurshid K. Recognition of printed Urdu ligatures using convolutional neural networks. Journal of Electronic Imaging. 2019 May;28(3):033004.
Singh, Y.K., 2016. Finding Connected Components in a Gray Scale Image. ADBU Journal of Engineering Technology, 5(2).
Mostafavi SM, Kazerouni IA, Haddadnia J. Noise removal from printed text and handwriting images using coordinate logic filters. In2010 International Conference on Computer Applications and Industrial Electronics 2010 Dec 5 (pp. 160-164). IEEE.
Devi H. Thresholding: A Pixel-Level image processing methodology preprocessing technique for an OCR system for the Brahmi script. Ancient Asia. 2006 Dec 1;1.
Singh TR, Singh OI. Image Resizing with Enhancement Technique on DCT Domain.
Pavlidis T. A thinning algorithm for discrete binary images. Computer graphics and image processing. 1980 Jan 1;13(2):142-57.
Albawi S, Mohammed TA, Al-Zawi S. Understanding of a convolutional neural network. In2017 International Conference on Engineering and Technology (ICET) 2017 Aug 21 (pp. 1-6). Ieee.
Sherstinsky A. Fundamentals of recurrent neural network (RNN) and long short-term memory (LSTM) network. Physica D: Nonlinear Phenomena. 2020 Mar 1;404:132306.
Hochreiter S. The vanishing gradient problem during learning recurrent neural nets and problem solutions. International Journal of Uncertainty, Fuzziness and Knowledge-Based Systems. 1998 Apr;6(02):107-16.
Hecht-Nielsen R. Theory of the backpropagation neural network. In Neural networks for perception 1992 Jan 1 (pp. 65-93). Academic Press.
Narayan S. The generalized sigmoid activation function: Competitive supervised learning. Information sciences. 1997 Jun 1;99(1-2):69-82.
Ganai AF, Koul A. Projection profile based ligature segmentation of Nastaleeq Urdu OCR. In2016 4th International Symposium on Computational and Business Intelligence (ISCBI) 2016 Sep 5 (pp. 170-175). IEEE.
Downloads
Published
How to Cite
Issue
Section
License
This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.
All papers should be submitted electronically. All submitted manuscripts must be original work that is not under submission at another journal or under consideration for publication in another form, such as a monograph or chapter of a book. Authors of submitted papers are obligated not to submit their paper for publication elsewhere until an editorial decision is rendered on their submission. Further, authors of accepted papers are prohibited from publishing the results in other publications that appear before the paper is published in the Journal unless they receive approval for doing so from the Editor-In-Chief.
IJISAE open access articles are licensed under a Creative Commons Attribution-ShareAlike 4.0 International License. This license lets the audience to give appropriate credit, provide a link to the license, and indicate if changes were made and if they remix, transform, or build upon the material, they must distribute contributions under the same license as the original.
