An optimized CNN system to recognize handwritten characters in ancient documents in Grantha script

An optical character recognition (OCR) system plays an important role in the digitization of ancient handwritten text document. Various adversaries of ancient documents such as ink stains, faded portion of text, humidity spots, and similar-shaped characters make the task of character recognition cha...

Celý popis

Uloženo v:
Podrobná bibliografie
Vydáno v:International journal of information technology (Singapore. Online) Ročník 15; číslo 4; s. 1975 - 1983
Hlavní autoři: Jindal, Amar, Ghosh, Rajib
Médium: Journal Article
Jazyk:angličtina
Vydáno: Singapore Springer Nature Singapore 01.04.2023
Springer Nature B.V
Témata:
Artificial Intelligence
Artificial neural networks
Computer Imaging
Computer Science
Convolution
Documents
Handwriting recognition
Image Processing and Computer Vision
Machine Learning
Neural networks
Optical character recognition
Optimization techniques
Original Research
Pattern Recognition and Graphics
Software Engineering
Vision
India
Optimized CNN
Character recognition
Grantha script
Ancient handwritten documents
ISSN:2511-2104, 2511-2112
On-line přístup:Získat plný text
Tagy: Přidat tag
Žádné tagy, Buďte první, kdo vytvoří štítek k tomuto záznamu!
Popis
Shrnutí:An optical character recognition (OCR) system plays an important role in the digitization of ancient handwritten text document. Various adversaries of ancient documents such as ink stains, faded portion of text, humidity spots, and similar-shaped characters make the task of character recognition challenging and tedious. This research study proposes an optimized convolution neural network (CNN) based OCR system to recognize each and every character present in the ancient document handwritten in Grantha script. A set of convolutional layers present in the proposed system extracts the deep hierarchical feature vectors from the input character image. Two fully connected neural network (FCNN) layers have classified these feature vectors into its correct class. The values of the hyper-parameters of CNN architecture such as the number of filters in each convolution layer, size of the filter in each convolution layer, number of FCNN layers, and neurons in each FCNN layer have been optimized using the Bayesian optimization technique. The major contribution of this work is the proposal of an optimized CNN architecture to perform OCR in ancient documents in Grantha script. A character recognition accuracy of 99.30% has been obtained from the proposed OCR system on the ancient handwritten documents in Grantha script. The experimental results demonstrate that the proposed OCR method outperforms the existing state-of-the-art methods in this regard.
Bibliografie:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
ISSN:2511-2104
2511-2112
DOI:10.1007/s41870-023-01247-1