Indian Scripts Research Articles

Offline Handwritten Gurumukhi Character Recognition System Using Convolution Neural Network

A lot of literature is available on the recognition of handwriting on scripts other than Indians, but the number of articles related to Indian scripts recognition such as Gurumukhi are much less. Gurumukhi is a religion-specific language that ranks 14th frequently spoken language in all languages of the world. In Gurumukhi script, some characters are alike to each other which makes recognition task very difficult. Therefore this article presents a novel approach for Gurumukhi character. This article lays emphasis on convolutional neural networks (CNN), which intend to obtain the features of given data samples and then its mapping is being performed to the right observation. In this approach, a dataset has been prepared for 10 Gurumukhi characters. The proposed methodology obtains a recognition accuracy of 99.34% on Gurumukhi characters images without making use of any post-processing method.

Multiobjective optimization for recognition of isolated handwritten Indic scripts

Identifying the most informative regions of character images is important for a robust handwritten character recognition system. If it is to be done in an efficient and cost-effective way, this task becomes more challenging. In this work, we have proposed a new multi-objective optimization framework to identify those local regions cost-effective way. We define three objective functions to be taken into account for this purpose: 1) recognition accuracy. 2) average recognition time per character image, and 3) redundancy of the local regions. More specifically, in this work, a modified opposition-based multiobjective Harmony Search algorithm has been proposed to select the local regions from handwritten character images based on their rankings in a three-dimensional pareto-front. Our work has been evaluated on four datasets - 1) isolated handwritten Bangla Basic characters, 2) isolated handwritten Bangla numerals, 3) English numerals, and 4) isolated handwritten Devanagari characters. The results show a significant decrease in the recognition costs and redundancy and an increase in the recognition accuracy obtained on all the datasets. Thus, our system provides a cost-effective approach towards isolated handwritten character recognition.

Hidden Markov Model Based Odia Numeral Recognition Using Moment and Structural Features

Optical character recognition (OCR) is a strategy to perceive character from optically checked and digitized pages. OCR plays an important role for Indian script research. The official language of the state Odisha is Odia. OCR face an incredible difficulties to recognize Odia language due to similar shape characters, their complex nature, the complicated way in which they combine form to compound character, use of Matra etc. Each character and numbers are passed through several modules like binarization, noise removal, segmentation, line segmentation, word segmentation, skeletonization, deskewing, thinning, thickening. The input picture is standardized to a size of 50 x 50 2D pictures. HMM is a stochastic process which has utilized in various applications for example speech recognition, Handwriting recognition, Gesture recognition. In this paper we utilized HMM to recognize the Odia character and numbers. Hidden Markov Model have many advantages such as resistant to noise, handle contrast recorded as a hard copy and the HMM devices are effectively accessible. In our proposed method we have developed an efficient recognition algorithm using Hidden Markov model based on moment based and structural feature to recognize Odia characters and numerals.

HWNet v2: an efficient word image representation for handwritten documents

We present a framework for learning an efficient holistic representation for handwritten word images. The proposed method uses a deep convolutional neural network with traditional classification loss. The major strengths of our work lie in: (i) the efficient usage of synthetic data to pre-train a deep network, (ii) an adapted version of the ResNet-34 architecture with the region of interest pooling (referred to as HWNet v2) which learns discriminative features for variable sized word images, and (iii) a realistic augmentation of training data with multiple scales and distortions which mimics the natural process of handwriting. We further investigate the process of transfer learning to reduce the domain gap between synthetic and real domain and also analyze the invariances learned at different layers of the network using visualization techniques proposed in the literature. Our representation leads to a state-of-the-art word spotting performance on standard handwritten datasets and historical manuscripts in different languages with minimal representation size. On the challenging iam dataset, our method is first to report an mAP of around 0.90 for word spotting with a representation size of just 32 dimensions. Furthermore, we also present results on printed document datasets in English and Indic scripts which validates the generic nature of the proposed framework for learning word image representation.

Indian Handwritten Script Identification System Based on Random Forest Tree Ensembles

The work proposal addresses to introduce a methodology for Indian unconstrained handwritten script identification by practicing distinct features and classifiers. By utilizing classifiers like RF, SVM, k-NN, and LDA for Indian script identification using statistical, geometric, and structural features. To preserve all the information present on handwritten documents such as historical, medieval, inscription, financial administration, public records, government archives, letters, land councils, various agreements, etc. in digitalize form needs textual document processing system (e.g. OCR). To build a precise and productive multi-script/language textual document processing system must have script identification. For this study use, total 1288 (line wise) samples of ten scripts use in India are collected from different persons of different gender, age, education and region (rural or urban). After successful training and testing, 81.8% and 0.252 accuracies and the OOB error rate are achieved by Random Forest respectively. And 77.8%, 73.5%, and 65.5% accuracy is achieved in SVM, k-NN and LDA classifiers respectively.

A clustering‐based feature selection framework for handwritten Indic script classification

AbstractIn India, which has numerous officially recognized scripts, there is a primary need for categorizing the documents on the basis of the scripts used therein. Identification of script used in a document is essential for its effective handling both manually and digitally. Identification of script in a document image is an important research problem in the pattern recognition field, which, at times, suffers from the issue of growing dimensionality of the feature vector and requires an efficient feature selection technique. Keeping this fact in mind, in this paper, we propose a clustering‐based filter feature selection framework in order to extract an optimal and effective feature subset from the original feature vector. The present feature selection methodology is evaluated on a script classification problem involving handwritten documents in 12 major Indic scripts. Experiments are done at word‐level, text‐line‐level, and block‐level. Experiments demonstrate that a reasonable increment in classification accuracy has been realized using comparatively lesser number of features. The proposed framework for feature selection is computationally inexpensive and can be applied to other pattern recognition problems as well.

An approach based on classifier combination for online handwritten text and non-text classification in Devanagari script

In this article, a method of analysing features of elliptical regions and combining outcomes of classifiers using Dempster–Shafer Theory (DST) is presented to classify online handwritten text and non-text data of any online handwritten document in the most popular Indic script—Devanagari. Although a few works exist in this regard in different non-Indic scripts, to our knowledge, no study is available to classify handwritten text and non-text document in online mode in any Indic script. The present method uses various structural and directional features analysed in elliptical regions to extract feature values from strokes of text and non-text data. The features are then studied separately in classification platforms based on Support Vector Machine (SVM) and Hidden Markov Model (HMM). The probabilistic outcomes of these two classification platforms are then combined using DST to improve the system performance. The efficiency of the present system has been measured on a self-generated dataset and it provides promising result.

Line Segmentation of Devanagari Ancient Manuscripts

In this paper, we have proposed a line segmentation algorithm for Devanagari ancient documents by dividing the document image into vertical stripes and followed by piecewise horizontal projection profiles. Average line height is computed, and based on this height, under-segmentation or over-segmentation of lines is detected for ancient manuscripts recognition. For experimental work of line segmentation algorithm, Devanagari ancient manuscripts are collected from various libraries and museums. To digitize these images, scanner and digital camera are used. The proposed line segmentation algorithm is tested on 1500 text lines. To the best of our knowledge, this is the first work of its kind for line segmentation on Devanagari ancient manuscripts. The proposed algorithm can be used for other Indian scripts also.

Feature Extraction Techniques for Text Recognition of Indian Scripts

Feature Extraction Techniques for Text Recognition of Indian Scripts

A survey of mono- and multi-lingual character recognition using deep and shallow architectures: indic and non-indic scripts

The cultural and regional diversity across the world and specifically in India has given birth to a large number of writing systems and scripts having a variety of character sets. For scripts having a larger character set, just a simple keyboard with limited character set is not the optimal way for providing inputs to the computer. Variations in individual handwriting due to mood swings, changes in medium of writing, changes in writing styles, etc. pose a challenge before the character recognition (CR) research community. Similar kinds of symbols in various scripts and languages act as a big barrier in multilingual CR. Lack of benchmark results and corpora for multilingual CR hinder the research in multilingual CR. There have been only a limited number of articles for optimal combination of features and classifiers to process multilingual data. Multilingual CR has least explored the Indic scripts. This paper presents a detailed review and analysis of the work done in multilingual online as well as offline CR for Indic and non-Indic scripts. The paper mainly contributes in two ways: Firstly, it provides a clear perspective about various phases of monolingual and multilingual CR; and secondly, identifies the major deficiencies in monolingual and multilingual CR for printed and handwritten text. It contributes by giving an in-depth view of work done at each phase including data acquisition, pre-processing, segmentation, feature extraction, recognition and post-processing of CR. Issues to be resolved at each phase have also been elaborated. The recent work done using Deep and Shallow architectures has been analysed. Tools used for these architectures have been compared to highlight their pros and cons. The present work also suggests how further research can be conducted in the field of monolingual and multilingual CR. The problems such as CR in hybrid documents, identifying more reliable features, resolving issues of similar characters, identifying optimal combination strategies for deep and shallow architectures, etc. need to be tackled in future research.

Understanding NFC-Net: a deep learning approach to word-level handwritten Indic script recognition

This paper presents a deep learning architecture modified for resource-constrained environments, called Non-Fully-Connected Network or NFC-Net, based on convolutional neural network architecture in order to solve the problem of Indic script recognition from handwritten word images. NFC-Net mainly targets resource constraint environment where there is a limited computation power or inadequate training samples or restricted training time. Previous approaches to handwritten script recognition included handcrafted features such as structure-based features and texture-based features. In contrast, here our model learns relatively different features from raw input pixels using NFC-Net. Various parameters of the NFC-Net are adjusted to present a vast and comprehensive study of the neural net in the domain of handwritten script recognition. In order to evaluate the performance of the NFC-Net with suitable parameter estimation, a dataset of 18,000 handwritten multiscript word images consisting of 1500 text words from each of the 12 officially recognized Indic scripts has been considered and a maximum script recognition accuracy of 96.30% is noted. Our proposed model also performs better than some of the recently published script recognition methods in bi-script, tri-script, tetra-script and 12-script scenarios. It has been additionally tested on the RaFD and BHCCD datasets with improved results to prove dataset independency of our model.

Mobile Based OCR Systems State-of-the-art Survey for Indian Scripts

International Journal of Computer Sciences and Engineering (A UGC Approved and indexed with DOI, ICI and Approved, DPI Digital Library) is one of the leading and growing open access, peer-reviewed, monthly, and scientific research journal for scientists, engineers, research scholars, and academicians, which gains a foothold in Asia and opens to the world, aims to publish original, theoretical and practical advances in Computer Science,Information Technology, Engineering (Software, Mechanical, Civil, Electronics & Electrical), and all interdisciplinary streams of Computing Sciences. It intends to disseminate original, scientific, theoretical or applied research in the field of Computer Sciences and allied fields. It provides a platform for publishing results and research with a strong empirical component. It aims to bridge the significant gap between research and practice by promoting the publication of original, novel, industry-relevant research.

Increasing the effectiveness of handwritten Manipuri Meetei-Mayek character recognition using multiple-HOG-feature descriptors

Detection and reading of the text from natural images is a difficult computer vision task, which is essential in a variety of emerging applications. Document character recognition is one such problem, which has been widely studied and documented by many machine learning and computer vision researchers, which is practically used for solving applications like recognizing handwritten digits. In this paper, a new approach for efficiently extracting cognition out of a total of 56 different classes of Handwritten Manipuri Meetei-Mayek (HMMM) (an Indian language) is described. Although character recognition algorithms have been researched and developed for other Indian scripts, no research work has been reported so far for recognizing all the characters of the Manipuri Meetei-Mayek (MMM). The work begins with a thorough analysis of the recognition task using a single hidden layer type Multilayer Perceptron Feedforward Artificial Neural Network with Histogram of Oriented Gradient (HOG) feature descriptors. After reviewing the level of accuracy and time it takes to train the network, the limitations are experimentally removed using multiple-sized cell grids using HOG descriptors. HOG, being a gradient-based descriptor, is very efficient in data discrimination and very stable with illumination variation. For efficient classification of the HOG features of the MMM, a linear multiclass support vector machine (SVM) classifier has been proposed for classifying the different offline characters because of its simplicity and speed. The classification based on linear multiclass SVM yielded a very high overall accuracy of 96.928%

RNN based online handwritten word recognition in Devanagari and Bengali scripts using horizontal zoning

Devanagari and Bengali scripts are two of the most popular scripts in India. Most of the existing word recognition studies in these two scripts have relied upon the widely used Hidden Markov Model (HMM), in spite of its familiar shortcomings. The existing works were evaluated against and performed well in their chosen metrics. But, the existing word recognition systems in these two scripts could not achieve more than 90% recognition accuracy. This article proposes a novel approach for online handwritten cursive and non-cursive word recognition in Devanagari and Bengali scripts based on two recently developed models of Recurrent Neural Network (RNN)—Long–Short Term Memory (LSTM) and Bidirectional Long–Short Term Memory (BLSTM). The proposed approach divides each word horizontally into three zones—upper, middle, and lower, to reduce the variations in basic stroke order within a word. Next, the word portions from middle zone are re-segmented into its basic strokes. Various structural and directional features are then extracted from each basic stroke of the word separately for each zone. These zone wise basic stroke features are then studied using both LSTM and BLSTM versions of RNN. Most of the existing word recognition systems in these two scripts have followed word based class labelling approach, whereas proposed system has followed the basic stroke based class labelling approach. An exhaustive experiment on large datasets has been performed to evaluate the performance of the proposed approach using both RNN and HMM to make a comparative performance analysis. Experimental results show that the proposed RNN based system is superior over HMM achieving 99.50% and 95.24% accuracies in Devanagari and Bengali scripts respectively and outperforms existing HMM based systems in the literature as well.

Handwritten Gujarati Character Recognition Using Structural Decomposition Technique

Handwritten character recognition is the active area of research. Development of Optical Character Recognition (OCR) system for Indian script like Gujarati is still in infancy and hence, there exists many unaddressed challenging problems for research community in this domain. The paper proposes three novel features to represent handwritten Gujarati characters. These features include features extracted based on structural decomposition, zone pattern matching and normalized cross correlation. Methods based on Support Vector Machine (SVM) and Naive Bayes (NB) classifiers have been exercised for the classification of Gujarati characters represented using proposed features. Experiments have been carried out on a dataset of 20500 handwritten Gujarati characters. Experimental results showed significant improvement over state-of-the-art when classifiers were learnt using structural decomposition based features.

DCWI: Distribution descriptive curve and Cellular automata based Writer Identification

Writer identification is an active area of research owing to its applications in a wide variety of fields, ranging from ancient document analysis to modern forensic document analysis. It deals with the writing style of documents and the learning of the discriminating features of different writers. In the domain of pattern recognition, the extraction of discriminative features of different writers has become very challenging. In order to address this concern, this work highlights a distribution descriptive curve (DDC-) and cellular automata (CA-) based model approach. The DDC utilizes the idea of the pixel distribution of handwritten text images to generate a unique curve as a feature vector. The generated feature vector is then fed to a support vector machine (SVM) as an input to identify the writer. Simultaneously, in a parallel mode, the initial handwritten text images are processed repeatedly with CA to generate another set of feature vectors. This new set of generated feature vectors are fed to a similarity-based classifier (SBC) as an input. The writer is predicted on the basis of the similarity of the features. The results from both of the approaches (DDC + SVM and CA + SBC) are merged to improve the performance of the model. This proposed model, DCWI, has good writer-identification capabilities compare with state-of-the-art techniques. Eventually, writer identification is accomplished using the ranking-based score scheme. The proposed model is evaluated on different datasets, e.g., IAM for English, IFN/ENIT for Arabic, and Kannada and Devanagari (Hindi) for the Indic script. The results show that the proposed model has better performance compared with existing state-of-the-art techniques.

Improved word-level handwritten Indic script identification by integrating small convolutional neural networks

Handwritten document recognition has been an active domain of research in the field of computer vision for several years since 1914 with the development of handheld scanner for reading printed texts called “optophone”. In India, which has several different scripts in one document page, identifying them is a must to automate process: document understanding. We propose a novel technique in integrating convolutional neural networks (CNNs) for script identification. We combined small individually trainable small CNNs, and used several different levels of variation in the architectures of the individual CNNs. Such a collection of individually trainable modules vary with respect to the input image size, CNN’s depth and wavelet transformation. In our test, we used publicly available dataset of size 11K words (1K per script) from 11 different Indic Scripts: Bangla, Devanagari, Gujarati, Gurumukhi, Kannada, Malayalam, Oriya, Roman, Tamil, Telugu and Urdu. Several ensemble strategies were implemented such as max-voting and probabilistic voting are used in addition to other conventional approaches like feature concatenation. We achieved a maximum accuracy of 95.04%, and it outperforms the accuracy of the state-of-the-art techniques like AlexNet by 2.9% and more importantly, benchmark techniques as (for script identification) on the dataset by more than 4%.

Tracing the Identity and Ascertaining the Nature of Brahmi-derived Devanagari Script

Current research exploits the orthographic design of Brahmi-derived scripts (also called Indic scripts), particularly the Devanagari script. Earlier works on orthographic nature of Brahmi-derived scripts fail to create a consensus among epigraphists, historians or linguists, and thus have been identified by various names, like semi-syllabic, subsyllabic, semi-alphabetic, alphasyllabary or abugida. On the contrary, this paper argues that Brahmi-derived scripts should not be categorized as scripts with overlapping features of alphabetic and syllabic properties as these scripts are neither alphabetic nor syllabic. Historical evolution and linguistic properties of Indic scripts, particularly Devanagari, ascertain the need for a new categorization of its own and, thus preferably merit a unique descriptor. This paper investigates orthographic characteristics of the Brahmi-derived Devanagari script, current trends in research pertaining to the Devanagari script along with other Indic scripts and the implications of these findings for literacy development in Indic writing systems.

Online Handwritten Gurmukhi Words Recognition

Identification of offline and online handwritten words is a challenging and complex task. In comparison to Latin and Oriental scripts, the research and study of handwriting recognition at word level in Indic scripts is at its initial phases. The two main methods of handwriting recognition are global and analytical. The present work introduces a novel analytical approach for online handwritten Gurmukhi word recognition based on a minimal set of words and recognizes an input Gurmukhi word as a sequence of characters. We employed a sequential step-by-step approach to recognize online handwritten Gurmukhi words. Considering the massive variability in online Gurmukhi handwriting, the present work employs the completely linked non-homogeneous hidden Markov model. In the present study, we considered the dependent, major-dependent, and super-dependent nature of strokes to form Gurmukhi characters in words. On test sets of online handwritten Gurmukhi datasets, the word-level accuracy rates are 85.98%, 84.80%, 82.40%, and 82.20% in four different modes. Besides the online Gurmukhi word recognition, the present work also provides Gurmukhi handwriting analysis study for varying writing styles and proposes novel techniques for zone detection and rearrangement of strokes. Our proposed algorithms have been successfully employed to online handwritten Gurmukhi word recognition in dependent and independent modes of handwriting.

Writing for Eternity: A Survey of Epigraphy in Southeast Asia

Writing for Eternity:A Survey of Epigraphy in Southeast Asia Anton O. Zakharov Daniel Perret Paris: École française d'Extrême-Orient, 2018, 478 pp. (Études thématiques 30). ISBN 978-2-85539-150-2; ISSN 1269-8067 Epigraphy is a key source for reconstructing history, at least in the Southeast Asian context. The only chronicles or other written sources left by many local polities are inscriptions engraved on stone or metal. History of epigraphic studies in Southeast Asia began more than two centuries ago, but there are still few truly seriously [End Page 142] examined corpora of various local inscriptions. A well-known exception is the fundamental corpus of inscriptions of Cambodia initiated by the French scholar George Cœdés: his seminal Inscriptions du Cambodge (1937–66) still remains the reference book for all who study the field. Another notable exception is Inscriptions of Burma (1933-6) by Gordon Hannington Luce and U Pe Maung Tin. In 2011 the French École française d'Extrême-Orient (EFEO) and the Ikatan Ahli Arkeologi Malaysia (Malaysian Archaeologists Association) launched a workshop 'Epigraphy of Southeast Asia' in Kuala Lumpur. After seven years, in 2018, the EFEO and Daniel Perret have published a comprehensive 'overview of epigraphy in Southeast Asia' (p. 7). Writing for Eternity offers a 'corpus' perspective on regional inscriptions and outlines the histories of epigraphic studies in Southeast Asia. The title of the book invokes the message of those who engraved the inscriptions on imperishable materials: the information preserved in them should survive forever, by grace of various gods, 'while the sun and the moon exist' as, for example, is claimed by the My Son inscription of King Sri Prakasadharman-Vikrantavarman C. 87 of 687 CE. Daniel Perret opens the volume with a substantial Introduction to the history of epigraphic researches in Southeast Asian states since Sir Thomas Stamford Raffles 'emphasise[d] the importance of inscriptions' for historical, linguistic and chronological studies of the Southeast Asian past (p. 14). Perret briefly overviews the corpora of inscriptions from Myanmar, Laos, Thailand, Cambodia, Campa, and maritime Southeast Asia that make use of Indian or Indian-derived scripts, as well as those that use Chinese or Chinese-derived, Arabic or Arabic-derived, Japanese, Armenian, Hebrew, and Latin scripts. Many languages of epigraphy are still in use; therefore, new inscriptions in them appear daily. Perret outlines questions for further research: why materials for engraving inscriptions differ in various countries; what role was played by the colonial administration for epigraphic studies; how epigraphy interacts with contemporary politics and/or nationality of scholars; the role played by private collectors and independent researchers in the field; how epigraphy interacts with other scientific disciplines like history, archaeology, religious studies etc. (p. 43). Strangely enough, Perret's Introduction does not mention Ramesh Candra Majumdar's Inscriptions of Champa (1927) and Inscriptions of Kambuja (1953). Despite many faults in his translations of Campa and Cambodian inscriptions, the Indian scholar did offer new readings and conjectures after Louis Finot, Édouard Huber and other French scholars had published the inscriptions of Campa for the first time. Thus, Majumdar translated the Mỹ Sơn inscription C. 81 of 712 CE which Finot left without translation. The volume contains three parts. The first part, 'Corpora of Inscriptions in Indian Scripts and Local Variants', outweighs the others in the diversity of papers and occupies around two–thirds of the volume. 'Myanmar Epigraphy: Current State and Future Tasks' by Tilman Frasch summarizes data on inscriptions from the country and their studies in several parts. The first deals with the inscriptions in languages other than Burmese, i.e. in the Pyu language, Sanskrit, Tamil and Pali. Many Pali texts were written on gold folios and were citations from the Pali Buddhist canon. This tradition was [End Page 143] well-established during the Pyu times in the first millennium CE. These Pali texts were placed inside Buddhist stupa and functioned as a substitute for the Buddha's corporal relics. Sanskrit inscriptions in Myanmar were largely issued by the kings of the Candra Dynasty of Vesali in Rakhine (Arakan). During the Pagan Kingdom, Pali served a lingua franca for Buddhist monks from Pagan (Bagan) itself, Sri...