Handwritten Devanagari Compound Character Recognition Using Legendre Moment: An Artificial Neural Network Approach

Kale, K. V.; Chavan, S. V.; Kazi, Majharoddin M.; Rode, Yogesh S.

doi:10.1109/iscbi.2013.62

Cited by 9 publications

(1 citation statement)

References 23 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…K.V. Kale et al [8], proposed a recognition system for handwritten devanagari Compound Character, based on Legendre moment feature descriptor. Moment function has been successfully applied to many pattern recognition problems.…”

Section: Literature Workmentioning

confidence: 99%

Unconstrained Character Classification For Off-Line Handwritten Devanagari Script Using Multi-Class Model For SVM

Kumar¹

2017

IJRASET

View full text Add to dashboard Cite

Hindi script is being used in various languages, for instance Marathi, Rajasthan, Sanskrit, and Nepali and it is also the script of Devanagari. It is observed that errors in classification mainly due to complex structures, incorrect segmentation and high unevenness in writing styles, classification of characters from the unconstrained script has become a burning vicinity of interest for researchers. Computer-based pattern recognition is a process that involves preprocessing, feature extraction, feature selection, and classification. In that article, we have extracted features from HOG, the novelty of this approach to attain better accuracy and reduce misclassification as well as for classification of handwritten characters with a multiclass model for SVM. Implementation has been performed using a self-created dataset of 40 users for handwritten Hindi characters. The experimental results obtained from this self-created dataset described the effectiveness of this system. The proposed system has faster speed and higher accuracy than the traditional Hindi OCR's. Enormous applications and the future necessities of optical character recognition area open new paths for researchers. An effort is made to address the most crucial consequences and it is also tried to foreground the better directions of the research till date. Our experimental results present the high performance of these features when classified using SVM classification.

show abstract

Section: Literature Workmentioning

confidence: 99%