Svm-pca based handwritten devanagari digit character recognition

被引:0
作者
Khamparia A. [1 ]
Singh S.K. [1 ]
Luhach A.K. [2 ]
机构
[1] Department of Computer Science and Engineering, Faculty of Technology and Sciences, Lovely Professional Universi-ty, Phagwara, Punjab
[2] Department of Electrical and Communication Engineering, Faculty of Technology and Sciences, The PNG University of Technology, Morobe, Papua New Guinea
关键词
Accuracy; Block based feature; Classifiers; Confusion; Devanagari script; Handwritten character recognition; Normalization; Prediction; Principal component analysis; Support vector machine;
D O I
10.2174/2213275912666181219092905
中图分类号
学科分类号
摘要
Objective: The blended fusion of Support Vector Machine (SVM) and Principal Component Analysis (PCA) have been widely used in recognizing handwritten digit characters of Devanagari script. The feature information from the character is extracted using its skeleton structure which optimally reduce data dimensionality using PCA. There is ample information available on handwritten charac-ter recognition on Indian and Non-Indian scripts but very few article emphasized on recognition of Devanagari scripts. Therefore, this paper presents an efficient handwritten Devanagari character recognition system based on block based feature extraction and PCA-SVM classifier. Methods: We have collected samples of handwritten Devanagari characters from different handwritten experts for classification. Results: For experimental work, total of 100 images having Devanagari digit characters been used for the purpose of training and testing. The proposed system achieves a maximum recognition accuracy of 96.6 % and 96.5% for 5 & 10 fold validations with 70% training and 30% testing data using block based feature and SVM classifier having different kernels. Conclusion: The obtained results achieve maximum accuracy using SVM classifier for digit character recognition. In future deep learning networks will be considered for accuracy enhancement and precision. © 2021 Bentham Science Publishers.
引用
收藏
页码:48 / 53
页数:5
相关论文
共 20 条
[1]  
Singh B., Mittal A., Ghosh D., An evaluation of different feature extractors and classifiers for offline handwritten devanagri character recognition, J. Pattern Recog. Res, 2, pp. 269-277, (2011)
[2]  
Wen Y., Lub Y., Shi P., Handwritten Bangla numeral recognition system and its application to postal automation, Patt. Recog, 40, 1, pp. 99-107, (2007)
[3]  
Tripathy J., Reconstruction of oriya alphabets using Zernike Mo-ments, Inter. J. Comp. App, 8, pp. 26-32, (2010)
[4]  
Jindal M. K., Degraded text recognition of gurmukhi script, (2008)
[5]  
Bhattacharya U., Shridhar M., Parui S. K., Sen P. K., Chaudhuri B. B., Offline recognition of handwritten Bangla characters an efficient two-stage approach, Patt. Ana. App, 15, 4, pp. 445-458, (2012)
[6]  
Kumar M., Jindal M. K., Sharma R. K., SVM based offline handwritten Gurmukhi character recognition, Proceedings of International Workshop on Soft Computing and Knowledge Dis-covery, 758, pp. 51-62, (2011)
[7]  
Kumar M., Jindal M. K., Sharma R. K., k-NN based offline handwritten Gurmukhi character recognition, Proceedings of International Conference on Information and Image Processing, pp. 1-4, (2011)
[8]  
Kumar M., Jindal M. K., Sharma, Classification of characters and grading writers in offline handwritten Gurumukhi script, Proceedings of International Conference on Information and Image Processing, pp. 1-4, (2011)
[9]  
Kumar M., Jindal M. K., Sharma R. K., Offline handwritten gurmukhi character recognition using curvature feature, Proceedings of International Conference on AMOC, pp. 981-989, (2011)
[10]  
Sundaram S., Ramakrishnan A. G., Two dimensional principal component analysis for online tamil character recognition, Proceedings of 11th International Conference Frontiers in Hand-writing Recognition, pp. 88-94, (2008)