Search

Optical Character Recognition OCR is an essential part of many real-world applications such as digital archiving, automatic number plate recognition, handle cheques, etc. However, developing an OCR for printed Arabic text is still a challenging and open research field due to the special characteristics of Arabic cursive script. In this paper, we propose a segmentation-based, omnifont, open-vocabulary OCR for printed Arabic text. The proposed approach doesn’t require an explicit font type recognition stage. It uses an explicit, indirect character segmentation method. The presented segmentation method is baseline dependent and employs a hybrid, three-steps character segmentation algorithm to handle the problem of character overlapping. Besides, it uses a set of topological features that are designed and generalized to make the segmentation approach font independent. The segmented characters are fed as an input to a convolutional neural network for feature extraction and recognition. The APTID-MF data set has been used for testing and evaluation. The average accuracy of the proposed segmentation stage is 95%, while the average accuracy of the recognition stage is 99.97%. The whole approach achieves an average accuracy of 95% without using font-type recognition or any post-processing techniques.

Recent advances in complex automated handwriting identification systems have led to a lack of understandability of these systems' computational processes and features by the forensic handwriting examiners that they are designed to support. To mitigate this issue, this research studied the relationship between two systems: FLASH ID® , an automated handwriting/black box system that uses measurements extracted from a static image of handwriting, and MovAlyzeR® , a system that captures kinematic features from pen strokes. For this study, 33 writers each wrote 60 phrases from the London Letter using cursive writing and handprinting, which led to thousands of sample pairs for analysis. The dissimilarities between pairs of samples were calculated using two score functions (one for each system). The observed results indicate that dissimilarity scores based on kinematic spatial-geometric pen stroke features (e.g., amplitude and slant) have a statistically significant relationship with dissimilarity scores obtained using static, graph-based features used by the FLASH ID® system. Similar relationships were observed for temporal features (e.g., duration and velocity) but not pen pressure, and for both handprinting and cursive samples. These results strongly imply that both the current implementation of FLASH ID® and MovAlyzeR® rely on similar features sets when measuring differences in pairs of handwritten samples. These results suggest that studies of biometric discrimination using MovAlyzeR® , specifically those based on the spatial-geometric feature set, support the validity of biometric matching algorithms based on FLASH ID® output.

According to statistics, there are 422 million speakers of the Arabic language. Islam is the second-largest religion in the world, and its followers constitute approximately 25% of the world’s population. Since the Holy Quran is in Arabic, nearly all Muslims understand the Arabic language per some analytical information. Many countries have Arabic as their native and official language as well. In recent years, the number of internet users speaking the Arabic language has been increased, but there is very little work on it due to some complications. It is challenging to build a robust recognition system (RS) for cursive nature languages such as Arabic. These challenges become more complex if there are variations in text size, fonts, colors, orientation, lighting conditions, noise within a dataset, etc. To deal with them, deep learning models show noticeable results on data modeling and can handle large datasets. Convolutional neural networks (CNNs) and recurrent neural networks (RNNs) can select good features and follow the sequential data learning technique. These two neural networks offer impressive results in many research areas such as text recognition, voice recognition, several tasks of Natural Language Processing (NLP), and others. This paper presents a CNN-RNN model with an attention mechanism for Arabic image text recognition. The model takes an input image and generates feature sequences through a CNN. These sequences are transferred to a bidirectional RNN to obtain feature sequences in order. The bidirectional RNN can miss some preprocessing of text segmentation. Therefore, a bidirectional RNN with an attention mechanism is used to generate output, enabling the model to select relevant information from the feature sequences. An attention mechanism implements end-to-end training through a standard backpropagation algorithm.

Urdu is one of the most popular languages in the world. It is a Persianized standard register of the Hindi language with considerable and valuable literature. While digital libraries are constantly replacing conventional libraries, a vast amount of Urdu literature is still handwritten. Digitizing this handwritten literature is essential to preserve it and make it more accessible. Nevertheless, the scarcity of Urdu Optical Character Recognition (OCR) research limits a digital library's scope to a manual document search. The limited research work in this area is mainly due to the complexity of Urdu Script. Unlike the English language, the Urdu writing style is cursive, bidirectional, and character shapes and sizes highly vary depending on their position. Holistic word recognition is found to be a better solution among many other text segmentation techniques as it takes the complete word into account instead of segmenting it explicitly or implicitly. For this project, the data of five different Urdu words were collected for training and testing a convolutional neural network and 96% recognition accuracy was achieved.

Multi-oriented handwritten documents require additional preprocessing for segmentation and subsequent phases to work accurately in handwritten recognition systems. Skew correction is one such additional phase. Appearance of skew in multi-oriented Indian language based handwritten document is higher due to the presence of cursive nature. In the current work, we utilise a salient feature present in Indian scripts called $m\bar {a}$ tr $\bar {a}$ (also known as headline), extract a group of eligible pixels, and employ linear curve fitting for detecting and correcting skew in handwritten words. The proposed method is capable of correcting skew in four distinct Indian languages, viz. Bangla, Hindi, Marathi, and Punjabi. It is capable of efficiently handling skewed word images to an extent of ± 55∘ and delivers precise result even when the $m\bar {a}$ tr $\bar {a}$ is mostly absent or discontinuous.

An optical character recognition (OCR) system may be the solution to data entry problems for saving the printed document as a soft copy of them. Therefore, OCR systems are being developed for all languages, and Kurdish is no exception. Kurdish is one of the languages that present special challenges to OCR. The main challenge of Kurdish is that it is mostly cursive. Therefore, a segmentation process must be able to specify the beginning and end of the characters. This step is important for character recognition. This paper presents an algorithm for Kurdish character segmentation. The proposed algorithm uses the projection-based approach concepts to separate lines, words, and characters. The algorithm works through the vertical projection of a word and then identifies the splitting areas of the word characters. Then, a post-processing stage is used to handle the over-segmentation problems that occur in the initial segmentation stage. The proposed method is tested using a data set consisting of images of texts that vary in font size, type, and style of more than 63,000 characters. The experiments show that the proposed algorithm can segment Kurdish words with an average accuracy of 98.6%.

In the area of machine learning, different techniques are used to train machines and perform different tasks like computer vision, data analysis, natural language processing, and speech recognition. Computer vision is one of the main branches where machine learning and deep learning techniques are being applied. Optical character recognition (OCR) is the ability of a machine to recognize the character of a language. Pashto is one of the most ancient and historical languages of the world, spoken in Afghanistan and Pakistan. OCR application has been developed for various cursive languages like Urdu, Chinese, and Japanese, but very little work is done for the recognition of the Pashto language. When it comes to handwritten character recognition, it becomes more difficult for OCR to recognize the characters as every handwritten character’s shape is influenced by the writer’s hand motion dynamics. The reason for the lack of research in Pashto handwritten character data as compared to other languages is because there is no benchmark dataset available for experimental purposes. This study focuses on the creation of such a dataset, and then for the evaluation purpose, a machine is trained to correctly recognize unseen Pashto handwritten characters. To achieve this objective, a dataset of 43000 images was created. Three Feed Forward Neural Network models with backpropagation algorithm using different Rectified Linear Unit (ReLU) layer configurations (Model 1 with 1-ReLU Layer, Model 2 with 2-ReLU layers, and Model 3 with 3-ReLU Layers) were trained and tested with this dataset. The simulation shows that Model 1 achieved accuracy up to 87.6% on unseen data while Model 2 achieved an accuracy of 81.60% and 3% accuracy, respectively. Similarly, loss (cross-entropy) was the lowest for Model 1 with 0.15 and 3.17 for training and testing, followed by Model 2 with 0.7 and 4.2 for training and testing, while Model 3 was the last with loss values of 6.4 and 3.69. The precision, recall, and f-measure values of Model 1 were better than those of both Model 2 and Model 3. Based on results, Model 1 (with 1 ReLU activation layer) is found to be the most efficient as compared to the other two models in terms of accuracy to recognize Pashto handwritten characters.

The remarkable growth of texts both in online and offline is becoming a challenging issue which need exploration for further research. Diversities of regional and cultural changes have produced diverse languages as a source of communication. Variations of styles are existing for handwritten texts which is due to varying writing styles. The research area of text recognition is matured which has increased a number of directions in the area of research. A detail report of the existing literature is needed which can help practitioners and researchers to use the existing evidence and provide new solutions for identification of cursive languages and to optimize the ability of recognition for cursive text. For facilitating the researchers and practitioners by providing in-depth analysis of the existing literature, the proposed study provide a detail report through which researchers can get benefit of the literature and devise new solutions. This study is based on searching various popular libraries for identifying relevant materials associated with the proposed study.

Decomposition of a word into a set of appropriate pseudo-characters is a challenging task in case of a cursive script like Bangla. Segmentation-free approach bypasses the decomposition problem entirely and treats the handwritten word as an individual entity. From the literature, we found that the accuracy of handwritten Bangla cursive word recognition using segmentation-free approach is relatively low (accuracy hovers between 80% and 90%). In the current work, we aim to provide a threefold study on this particular domain. Firstly, we extract different statistical feature sets from word images and use five different off-the-shelf classifiers to delineate their performance. Then, we employ five different CNN-TL architectures, namely AlexNet, VGG-16, VGG-19, ResNet50, and GoogleNet, to understand how they perform on holistic Bangla words. Finally, we use a seven-layer FCN architecture and provide a comparison of results with all the aforementioned experimentations. We achieved an accuracy of 98.86% with ResNet50, which is nearly 19% improvement when compared with other recent state-of-the-art methodologies.

Signboard detection and recognition is an important task in automated context-aware marketing. Recently many scripting languages like Latin, Japanese, and Chinese have been effectively detected by several machine learning algorithms. As compared to other languages, outdoor Urdu text needs further attention in detection and recognition due to its cursive nature. Urdu detection and recognition are also difficult due to a wide variety of illuminations, low resolution, inconsistent font styles, color, and backgrounds. To overcome the deficiency of Urdu text detection from the outdoor environment, we have proposed a new Urdu-text signboard dataset with 467 ligature categories, containing a 30 + K images for recognition and 700 base images with annotation are created for detection. We also propose a methodology, that consists of 3-phases. In first phase text regions containing Urdu ligatures from shop-signboard images are detected by a faster regional convolutional neural network (FasterRCNN) using pre-trained CNNs like Alexnet and Vgg16. In the second phase detected regions from the first phase are clustered to identify unique ligatures in a dataset. Lastly in the third phase, all detected regions are recognized by 18-layer convolutional neural network trained model. The proposed system has successfully achieved the precision and recall of 87% and 96% respectively using vgg16 model for detection. For the classification of ligatures, a recognition rate of 97.50% is achieved. Recognition of ligatures was also evaluated using bilingual evaluation understudy (BLEU), and achieved an encouraging score of 0.96 on the newly developed Urdu-Signboard dataset.

Optical character recognition, known as OCR, has been widely used due to high demand of different technologies. Currently, most existing OCR systems have been focused on Latin languages. In recent studies, OCR systems for non-Latin texts involving cursive style have also been introduced despite posing some challenges. In this study, the authors propose an OCR system based on long short-term memory neural networks for the Persian language. The authors also investigate the effects of variations of parameters, involved in this approach. The proposed OCR system solves false recognition of sub-word ‘LA’ and ‘LA’. Moreover, the authors present a preprocessing algorithm to remove ‘justification’ using image processing. A new comprehensive collated data set is introduced, comprising five million images with eight popular Persian fonts and in ten various font sizes. The proposed evaluations show that the accuracy of the proposed OCR is increased by 2%, compared to the existing Persian OCR system. The experimental results indicated that the proposed system has average accuracy of 99.69% at the letter level. The proposed system has an accuracy of 98.1% for ‘zero-width non-breaking space’ and 98.64% for ‘LA’ at the word level.

This study focuses on cursive text recognition appearing in videos, using a complete framework of deep neural networks. While mature video optical character recognition systems (V-OCRs) are available for text in non-cursive scripts, recognition of cursive scripts is marked by many challenges. These include complex and overlapping ligatures, context-dependent shape variations and presence of a large number of dots and diacritics. The authors present an analytical technique for recognition of cursive caption text that relies on a combination of convolutional and recurrent neural networks trained in an end-to-end framework. Text lines extracted from video frames are preprocessed to segment the background and are fed to a convolutional neural network for feature extraction. The extracted feature sequences are fed to different variants of bi-directional recurrent neural networks along with the ground truth transcription to learn sequence-to-sequence mapping. Finally, a connectionist temporal classification layer is employed to produce the final transcription. Experiments on a data set of more than 40,000 text lines from 11,192 video frames of various News channel videos reported an overall character recognition rate of 97.63%. The proposed work employs Urdu text as a case study but the findings can be generalised to other cursive scripts as well.

Motivation : The problem of handwritten text recognition is vastly studied since last few decades. Many innovative ideas have been developed, where state-of-the-art accuracy is achieved for the English, Chinese or Indian scripts.The recent developments for the cursive scripts such as Arabic and Urdu handwritten text recognition have achieved remarkable accuracy. However, for the Sindhi script, existing systems have not shown significant results and the problem is still an open challenge. Several challenges such as variations in writing styles, joined text, ligature overlapping, and others associated to the handwritten Sindhi text make the problem more complex. Objectives: In this study, a deep residual network with shortcut connections and summation fusion method using convolutional neural network (CNN) is proposed for automatic feature extraction and classification of handwritten Sindhi characters. Method: To increase the powerful feature representation ability of the network, the features of the convolutional layers in the residual block are fused together and combined with the output of the previous residual block. The proposed network is trained on a custom developed handwritten Sindhi character dataset. To tackle the problem of small data, a data augmentation with rotation, flipping and image enhancement techniques have been used. Findings: The experimental results show that the proposed model outperforms than the best results previously published for the handwritten Sindhi character recognition. Novelty: This is the first research that proposes deep residual network with summation fusion for the Sindhi handwritten text recognition. Keywords: Handwritten Sindhi character recognition; Sindhi text recognition; cursive text recognition; deep learning; ResNet; convolutional neural network

Chinese is a widely used language in the world. Chinese cursive script is one of the most distinctive calligraphy art and traditional cultures of China. However, for its connected writing, there is a lack of research on text recognition for cursive images. Here, the authors construct a small cursive image dataset named as Chinese Cursive and there are 523 images in this dataset. It contains continuous strokes text, recognises difficulty etc. Each cursive character is corresponded to a label. The authors proposed a cursive detection method named as SE-seglink for the dataset. The SE-seglink further enhances the image feature extraction. Compared to the existing methods, the SE-seglink performs better in recognising cursive scripts and improves the precision of text detection in cursive images. After multiple sets of comparative experiments, the effectiveness of the SE-seglink method was evaluated by the experiment on the benchmark image dataset ICDAR2015.

Building a robust Optical Character Recognition (OCR) system for languages, such as Arabic with cursive scripts, has always been challenging. These challenges increase if the text contains diacritics of different sizes for characters and words. Apart from the complexity of the used font, these challenges must be addressed in recognizing the text of the Holy Quran. To solve these challenges, the OCR system would have to undergo different phases. Each problem would have to be addressed using different approaches, thus, researchers are studying these challenges and proposing various solutions. This has motivate this study to review Arabic OCR dataset because the dataset plays a major role in determining the nature of the OCR systems. State-of-the-art approaches in segmentation and recognition are discovered with the implementation of Recurrent Neural Networks (Long Short-Term Memory-LSTM and Gated Recurrent Unit-GRU) with the use of the Connectionist Temporal Classification (CTC). This also includes deep learning model and implementation of GRU in the Arabic domain. This paper has contribute in profiling the Arabic text recognition dataset thus determining the nature of OCR system developed and has identified research direction in building Arabic text recognition dataset.

In this paper, we propose integrated adaptive sensitivity and robustness terms for the cost function of a convolutional neural network (CNN). The sensitivity term considers the slight variations and high frequency components of the input image samples. It distinguishes between images that look similar but belong to different classes. This regularizer is designed to enhance the between-class distance which is a biological definition for the simple cells of the visual system. On the other hand, the robustness term is used to develop a more stable CNN structure against disturbances and perturbations. The robust term provides better within-class features because it recognizes images that look different but are actually from the same class. The robust term symbolizes the complex cell characteristics of the visual system. The coefficients of the sensitivity and robustness regularization terms are adaptively tuned along with the network parameters using gradient descent. Two optimizers are assigned to tune the parameters: one for tuning the model parameters and the other one to adjust the sensitivity and robustness coefficients. This approach is applied to Korean traditional documents for character classification. The results show better within- and between-class classification ability for highly complex character styles with imbalanced number of samples.

Recognition of cursive handwritten Arabic text is a difficult problem because of context-sensitive character shapes, the non-uniform spacing between words and within a word, diverse placements of dots, and diacritics, and very low inter-class variation among individual classes. In this paper, we review and investigate different deep learning architectures and modeling choices for Arabic handwriting recognition. Further, we address the problem that imbalanced data sets present to deep learning systems. In order to address this issue, we are presenting a novel adaptive data-augmentation algorithm to promote class diversity. This algorithm assigns a weight to each word in the database lexicon. This weight is calculated based on the average probability of each class in a word. Experimental results on the IFN/ENIT and AHDB databases have shown that our presented approach yields state-of-the-art results.

Urdu text is a cursive script and belongs to a non-Latin family of other cursive scripts like Arabic, Chinese, and Hindi. Urdu text poses a challenge for detection/localization from natural scene images, and consequently recognition of individual ligatures in scene images. In this paper, a methodology is proposed that covers detection, orientation prediction, and recognition of Urdu ligatures in outdoor images. As a first step, the custom FasterRCNN algorithm has been used in conjunction with well-known CNNs like Squeezenet, Googlenet, Resnet18, and Resnet50 for detection and localization purposes for images of size $320\times 240$ pixels. For ligature Orientation prediction, a custom Regression Residual Neural Network (RRNN) is trained/tested on datasets containing randomly oriented ligatures. Recognition of ligatures was done using Two Stream Deep Neural Network (TSDNN). In our experiments, five-set of datasets, containing 4.2K and 51K Urdu-text-embedded synthetic images were generated using the CLE annotation text to evaluate different tasks of detection, orientation prediction, and recognition of ligatures. These synthetic images contain 132, and 1600 unique ligatures corresponding to 4.2K and 51K images respectively, with 32 variations of each ligature (4-backgrounds and font 8-color variations). Also, 1094 real-world images containing more than 12k Urdu characters were used for TSDNN’s evaluation. Finally, all four detectors were evaluated and used to compare them for their ability to detect/localize Urdu-text using average-precision (AP). Resnet50 features based FasterRCNN was found to be the winner detector with AP of.98. While Squeeznet, Googlenet, Resnet18 based detectors had testing AP of.65,.88, and.87 respectively. RRNN achieved and accuracy of 79% and 99% for 4k and 51K images respectively. Similarly, for characters classification in ligatures, TSDNN attained a partial sequence recognition rate of 94.90% and 95.20% for 4k and 51K images respectively. Similarly, a partial sequence recognition rate of 76.60% attained for real world-images.

One of the most challenging module in any handwritten character recognition system is separating individual characters from the document. Better character recognition rate is highly influenced by how much accuracy, character segmentation phase, has achieved. As the input data set is contains documents written in MODI script, and since MODI script characters are written in cursive manner, obtaining individual characters from the text becomes extremely complicated. Moreover, there is no separation between two adjacent words. Although now-a-days, Marathi language uses Devnagari script, several years back MODI Script was used for writing documents in Marathi language. This work presented in this paper tackles segmentation of individual characters from ancient handwritten MODI Script documents. Vertical Projection Profile (VPP) method can segment characters from a given line accurately only when two adjacent characters are separated by a zero-pixel column. Since characters of one line in this script are written without lifting hand, output of VPP does not suffice. This paper presents a novel approach for isolating individual characters from the line obtained using authors’ previous research work which uses dual thresholding criterion to minimize the character segmentation error. The techniques used in this paper are straightforward and reasonably economical in terms of execution time efficiency. The results obtained are quite satisfactory.

In this study, we consider a problem of modeling and manipulating hairy-brush characters based on the so-called ‘dynamic font’ method with control-theoretic B-spline approach, in which the characters are constituted as a result of trajectory curves using normalized uniform B-splines as basis function. First, typefaces of pre-designed dynamic font characters with one-pixel thickness are transformed to hairy-brush ones by introducing a deep learning method ‘Pix2Pix’. We then develop a method for modeling hairy-brush characters by formulating the problem as an optimal function-approximation problem, which minimizes an input energy of writing trajectory. Also, manipulations for generating cursive words as seen in Japanese calligraphy in a systematic way are described. A design example of cursive words is included.

The rise of Artificial Intelligence technology along with machine and deep learning are opening up almost limitless possibilities. There is also an element of fear towards this exponential growth. But it’s the humanization of our machines and devices that is seldom discussed or reported. But we deem it acceptable to nominate an unfortunate volunteer to retype notes scribbled on tiny pieces of paper so they can send to other attendees. Once again we need to ask ourselves if this is the most efficient method of managing our workload in this digital age. In order to solve this problem, we are proposing a solution to solve this. Instead of predicting by word, here we will be segregating the cursive English letter to individual characters and predicting it via trained Convolutional Neural Network (CNN) model. By using this methodology, increase in rate of prediction of the individual characters is been increased and it could be implemented to digitalize the forms in companies which will be a greater level of automation.

The accuracy of current natural scene text recognition algorithms is limited by the poor performance of character recognition methods for these images. The complex backgrounds, variations in the writing, text size, orientations, low resolution and multi-language text make recognition of text in natural images a complex and challenging task. Conventional machine learning and deep learning-based methods have been developed that have achieved satisfactory results, but character recognition for cursive text such as Arabic and Urdu scripts in natural images is still an open research problem. The characters in the cursive text are connected and are difficult to segment for recognition. Variations in the shape of a character due to its different positions within a word make the recognition task more challenging than non-cursive text. Optical character recognition (OCR) techniques proposed for Arabic and Urdu scanned documents perform very poorly when applied to character recognition in natural images. In this paper, we propose a multi-scale feature aggregation (MSFA) and a multi-level feature fusion (MLFF) network architecture to recognize isolated Urdu characters in natural images. The network first aggregates multi-scale features of the convolutional layers by up-sampling and addition operations and then combines them with the high-level features. Finally, the outputs of the MSFA and MLFF networks are fused together to create more robust and powerful features. A comprehensive dataset of segmented Urdu characters is developed for the evaluation of the proposed network models. Synthetic text on the patches of images with real natural scene backgrounds is generated to increase the samples of infrequently used characters. The proposed model is evaluated on the Chars74K and ICDAR03 datasets. To validate the proposed model on the new Urdu character image dataset, we compare its performance with the histogram of oriented gradients (HoG) method. The experimental results show that the aggregation of multi-scale and multilevel features and their fusion is more effective, and outperforms other methods on the Urdu character image and Chars74K datasets.

This paper demonstrates the importance of various pre-processing steps for the noisy character recognition system with the help of MATLAB simulated results. Image pre-processing is one of the first and foremost steps for working with images of noisy characters among handwritten characters, printed characters, online written characters or offline written character, and it can have mixed font characters, omni-font characters, boxed discrete character, spaced discrete character or pure cursive script writing. Pre-processing of such image aims at image noise reduction, image normalization, and image compression. This paper shows that there are various steps/methods of image pre-processing. The sequence and use of a particular step/method is used based on the application requirement. The simulation result presented in this paper highlights the usability of each pre-processing steps and review the commonly used methods of each preprocessing stage.

Processing and analyzing historical manuscripts is considered one of the most challenging problems in the document analysis and recognition domain. Manuscripts written in cursive are even more difficult due to overlapping words with random spacing, irregular and varying characters' shapes, poor scan quality, and insufficient labeled data. Despite the significant achievements of deep learning approaches in computer vision, handwritten word recognition is far from solved. Most of the existing methods focus on well-segmented word datasets. In this paper, we present an empirical study investigating how well state-of-the-art deep learning models perform on detection and recognition of handwritten words in Spanish American notary records. Professional historians were involved in preparing a labeled dataset of 26,482 Spanish words employed in the experiments. We investigate the performance of some state-of-the-art models on optical character recognition (OCR) on handwritten text documents: Keras-OCR, the object detection algorithm "You Only Look Once" (YOLO), Tesseract OCR, Kraken, and Calamari-OCR. Since YOLO does not include a text recognizer, we propose YOLO-OCR, an innovative model to detect and recognize words in historical manuscripts written in Spanish. Our results show the performance of pre-trained models on our dataset and that Keras-OCR and YOLO-OCR models are highly valuable for content extraction.

Pashto is one of the most ancient and historical languages in the world and is spoken in Pakistan and Afghanistan. Various languages like Urdu, English, Chinese, and Japanese have OCR applications, but very little work has been conducted on the Pashto language in this perspective. It becomes more difficult for OCR applications to recognize handwritten characters and digits, because handwriting is influenced by the writer’s hand dynamics. Moreover, there was no publicly available dataset for handwritten Pashto digits before this study. Due to this, there was no work performed on the recognition of Pashto handwritten digits and characters combined. To achieve this objective, a dataset of Pashto handwritten digits consisting of 60,000 images was created. The trio deep learning Convolutional Neural Network, i.e., CNN, LeNet, and Deep CNN were trained and tested with both Pashto handwritten characters and digits datasets. From the simulations, the Deep CNN achieved 99.42 percent accuracy for Pashto handwritten digits, 99.17 percent accuracy for handwritten characters, and 70.65 percent accuracy for combined digits and characters. Similarly, LeNet and CNN models achieved slightly less accuracies (LeNet, 98.82, 99.15, and 69.82 percent and CNN, 98.30, 98.74, and 66.53 percent) for Pashto handwritten digits, Pashto characters, and the combined Pashto digits and characters recognition datasets, respectively. Based on these results, the Deep CNN model is the best model in terms of accuracy and loss as compared to the other two models.

Although numerous research studies have been conducted concerning automatic vehicle number plate detection and recognition, various presented automated number plate recognition systems are devised for specific countries where number plates follow standard patterns. However, such systems cannot be applied in Afghanistan because of the different designs and the language. Moreover, due to the cursive nature, writing direction, and shape variation of the Pashto characters, the segmentation of words into isolated characters is a more complicated task. Hence, the Pashto optical character recognition is a less developed area. To date, no research study has been conducted for Afghanistan number plate detection and recognition. The details on the Afghanistan number plate include character, numbers, and each province's name. The paper presents the study of its type attempting to detect the number plate from the vehicle image and then recognize the province's name, characters, and numbers on the number plate. In particular, the new method incorporating four core steps. The first step is number plate detection applying canny edge detection based on user-defined thresholding and extracts the number plate involving several image processing techniques. The second phase is number plate adjustment using Randon transform-based techniques. The third stage is number plate segmentation isolating each character, number, and province name on the number plate using a scanning approach. The final step employs a convolutional neural network to classify the number plate's alphanumeric characters and provinces' names. In addition, two datasets have been created: the dataset for alphanumeric characters contains 2800 images of 14 classes, and the dataset for provinces' names contains 6800 images of 34 classes. The proposed models present 99.93 percent accuracy for provinces' names classification and 98.93 percent for alphanumeric characters' classification.

Inability to read doctors’ handwritten prescriptions causes 7,000 deaths a year in a developed country like the US. The situation should be worse in developing countries where more doctors use handwriting prescriptions. In Bangladesh, the writings become more indecipherable as they contain both English and Bangla words with Latin abbreviations of medical terms. As a result, patients and pharmacists find them difficult to read and the pharmacists provide wrong medicines. In order to ease the difficulty of reading doctors’ prescriptions, this paper proposes an online handwritten recognition system to predict the doctors’ handwriting and develop a digital prescription. To build this system, the “Handwritten Medical Term Corpus” dataset is introduced which contains 17,431 data samples of 480 words (360 English and 120 Bangla) from 39 Bangladeshi doctors and medical professionals. A bigger sample size can improve the recognition efficiency. A new data augmentation technique SRP (Stroke Rotation and Parallel shift) method is proposed to widen the variety of handwriting styles and increase the sample size. A sequence of line data is extracted from the augmented image dataset of 1,591,100 samples which is fed to a Bidirectional LSTM model. The proposed method has achieved 89.5% accuracy which is 16.1% higher than the recognition accuracy with no data expansion. This technology can reduce medical errors and save medical cost and ensure healthy living.

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.

© 2019 IEEE. The recognition of text in natural scene images is a practical yet challenging task due to the large variations in backgrounds, textures, fonts, and illumination. English as a secondary language is extensively used in Gulf countries along with Arabic script. Therefore, this paper introduces English-Arabic scene text recognition 42K scene text image dataset. The dataset includes text images appeared in English and Arabic scripts while maintaining the prime focus on Arabic script. The dataset can be employed for the evaluation of text segmentation and recognition task. To provide an insight to other researchers, experiments have been carried out on the segmentation and classification of Arabic as well as English text and report error rates like 5.99% and 2.48%, respectively. This paper presents a novel technique by using adapted maximally stable extremal region (MSER) technique and extracts scale-invariant features from MSER detected region. To select discriminant and comprehensive features, the size of invariant features is restricted and considered those specific features which exist in the extremal region. The adapted MDLSTM network is presented to tackle the complexities of cursive scene text. The research on Arabic scene text is in its infancy, thus this paper presents benchmark work in the field of text analysis.

In the emerging age of technologies, machines are becoming more and more skilled and capable just like humans. Despite the fact that machines do not have their own intelligence, but still due to advancement in Artificial Intelligence (AI), machines are rapidly advancing. The area of Pattern Recognition (PR) deals with bringing enhancements to identify obscure patterns corresponding to specific classes. Optical Character Recognition (OCR) is a subfield of PR which deals with the recognition of characters. A great work has been done for Japanese, Hindi, Arabic and Chinese scripts, but only a diminutive work has been done for Urdu script. The Urdu language is highly cursive and is written in different calligraphic styles like Naskh, Nastalique, Kofi, Devani and Riqa. The Nastalique font is very calligraphic with aesthetic beauty. The ligature segmentation of Urdu Nastalique is also more difficult as compared to other languages. Urdu Nastalique has some characteristics like stacking of ligatures and cursiveness which makes its ligature segmentation a difficult task. Cursiveness means ligatures are joined together to form a new shape. It contains connected ligatures which makes it more complicated as compared to other languages. The ligature recognition of Urdu text by an OCR is a strenuous task due to variants of scaling, rotation, orientation and font style. In this study, a scale and rotation invariant classifier for Urdu Nastalique OCR is proposed. A combination of scale and location invariant moments is used for feature extraction and the classification is performed using Cascade Forward Backpropagation Neural Network. The model is validated through independent dataset testing and 5-fold cross-validation which gave 96.474% and 96.922% accuracy. The results depict the adaptability of the proposed model due to its high accuracy for recognition of Urdu Nastalique Ligature.

The state-of-the-art Urdu recognition approaches for Nastalique use features along with the sequence of characters’ labels for classification and recognition. In Arabic-like cursive script, the characters are joined together to form a ligature. The conventional methods process the connected stroke of ligatures as a sequence of characters. However, connected stroke of a ligature image has a sequence of pairs of characters and their joiners, instead of a sequence of characters. The character has a distinctive shape that clearly distinguishes it from other characters. The joiner preserves the connecting stroke shape of a character with the next character. In this paper, an implicit Urdu character recognition technique is presented for the Nastalique writing style that is based on recognition of characters and joiners. The detailed analysis of the Nastalique calligraphy is carried out to extract the artistic features of characters and their joiners. The presented technique is tested on Dataset-1 of 1446 ligature classes covering 3309762 ligature instances and 91129 unique Urdu words. In addition, the system is also tested on 1600 text lines of UPTI dataset called Dataset-2. The character recognition accuracies are 95.58% and 98.37% on Dataset-1 and Dataset-2, respectively. The results reveal that the system outperforms the state-of-the-art hidden Markov models and deep learning-based Urdu recognition techniques.

Recognition of historical documents is a challenging problem due to the noised, damaged characters, and background. However, in Japanese historical documents, not only contains the mentioned problems, pre-modern Japanese characters were written in cursive and are connected. Therefore, character segmentation-based methods do not work well. This leads to the idea of creating a new recognition system. In this paper, we propose a human-inspired document reading system to recognize multiple lines of pre-modern Japanese historical documents. During the reading, people employ eyes movement to determine the start of a text line. Then, they move the eyes from the current character/word to the next character/word. They can also determine the end of a line or skip a figure to move to the next line. The eyes movement integrates with visual processing to operate the reading process in the brain. We employ attention-based encoder–decoder to implement this recognition system. First, the recognition system detects were to start a text line. Second, the system scans and recognize character by character until the text line is completed. Then, the system continues to detect the start of the next text line. This process is repeated until reading the whole document. As results, the system is successful to recognize multiple lines, connected and cursive characters without performing character/line segmentation. Besides, we also employ a coverage model which stores the history of eyes movement to predict the next movement more precisely. We tested our human-inspired recognition system on the pre-modern Japanese historical document provided by the PRMU Kuzushiji competition. The results of the experiments demonstrate the superiority and effectiveness of our proposed system by achieving Sequence Error Rate of 9.87% and 53.81% on level 2 and level 3 of the dataset, respectively. These results outperform to any other systems participated in the PRMU Kuzushiji competition.

The main problem of Bangla (Bengali) and Devanagari handwriting recognition is the shape similarity of characters. There are only a few pieces of work on writer-independent cursive online Indian text recognition, and the shape similarity problem needs more attention from the researchers. To handle the shape similarity problem of cursive characters of Bangla and Devanagari scripts, in this article, we propose a new category of features called ‘ sub-stroke-wise relative feature ’ (SRF) which are based on relative information of the constituent parts of the handwritten strokes. Relative information among some of the parts within a character can be a distinctive feature as it scales up small dissimilarities and enhances discrimination among similar-looking shapes. Also, contextual anticipatory phenomena are automatically modeled by this type of feature, as it takes into account the influence of previous and forthcoming strokes. We have tested popular state-of-the-art feature sets as well as proposed SRF using various (up to 20,000-word) lexicons and noticed that SRF significantly outperforms the state-of-the-art feature sets for online Bangla and Devanagari cursive word recognition.

It is a very challenging task to recognize unconstrained Bengali handwritten text due to its cursive nature. This paper introduces an offline technique of recognizing handwritten Bengali sentences based on BiLSTM architecture and connectionist temporal classification (CTC) output layer. The traditional approach of detecting handwritten sentence recognition rises the computational complexity due to the difficulty of combining isolated characters. The proposed method only segments the sentences into words, recognizes separately and combines them to represents the output sentence. The proposed technique is evaluated with 450 Bengali sentences and achieved 92% accuracy for recognizing sentences considering 15 distinct writing styles.

One of the major research areas in data science is deep learning. Convolutional Neural Networks play a huge role in the recent success of the artificial intelligence-based applications. Cursive character recognition is still an active research problem from many years. This paper deals with recognition of offline handwritten Marathi documents written in MODI script using deep learning convolutional neural network model. The algorithm explained in this paper makes use of Rectified Linear Unit neuron activation function in the convolutional layers as it gives better performance and computationally least expensive and softmax neuron activation function in its fully connected layer. The max pooling layer is used for feature detection. The model designed uses 32 convolutional filters of size 3 × 3 × 1. The model developed is trained by using more than 40000-character samples belonging to 31 different class labels. The trained neural network is tested on more than 10000-character samples from 31 different class labels, giving an accuracy of more than 60%.

In this paper, a robust character segmentation approach for cursive handwritten Modi script touching character cluster is presented. Prior to segmentation, the middle text region of the touching character cluster is separated by examining the location of Shirorekha and baseline. The middle text region is scrutinized for the estimation of ligature between two characters. Two different strategies are employed to find the location of the ligature. The selection of the strategy is based on the degree of connected component overlapratio. The foreground pixel intensity and vertical projection profile is scrutinized to segment the touching characters. The performance of the system is tested using the touching character clusters of the original archaic handwritten Modi documents. The proposed approach yields efficient touching characters cluster segmentation output and it is feasible to tackle most of the challenges in touching character cluster segmentation.

Through the utilization and convergence of technologies, such as artificial intelligence, big data, and the Internet of Things, that can represent the fourth industrial revolution, today’s world is changing rapidly. However, public organizations in Korea are still manually entering reports written by hand in many fields into the information system by a government official in charge. This study studies the improvement method for optical character recognition (OCR) of hand-written General Educational Development (GED) application. Through learning using cursive fonts, separating images for items subject to information service, and managing whitelist by item, the recognition rate is to be improved. In the field of optical character recognition (OCR), the recognition rate of printed texts written in Hangeul font is over 90%, but the recognition rate of cursive writing is very low. Various tables and letters of different sizes in hand-written application form documents hinder the recognition rate. Although the recognition rate has improved a lot recently with various studies using Deep Learning, there is a problem that takes too much processing time. Of the 11,172 Hangeul characters, 2350 are used. On the other hand, English has 52 case-sensitive characters. The recognition rate of Hangeul, which consists of a combination of initial sound, medial vowels, and final consonant, is bound to drop significantly. As mentioned earlier about a performance issue, software with open source-based neural network learning engines was used. In addition, the status of Hangeul used in names and addresses was investigated for the application of whitelists. The cursive font learning, image separation and whitelist proposed in this study have been effective in improving recognition rates.

Owing to the overwhelming accuracy of the deep learning method demonstrated at the 2012 image classification competition, deep learning has been successfully applied to a variety of other tasks. The high-precision detection and recognition of Kuzushiji, a Japanese cursive script used for transcribing historical documents, has been made possible through the use of deep learning. In recent years, competitions on Kuzushiji recognition have been held, and many researchers have proposed various recognition methods. This study examines recent research trends, current problems, and future prospects in Kuzushiji recognition using deep learning.

One of the most recent challenging issues of pattern recognition and artificial intelligence is Arabic text recognition. This research topic is still a pervasive and unaddressed research field, because of several factors. Complications arise due to the cursive nature of the Arabic writing, character similarities, unlimited vocabulary, use of multi-size and mixed-fonts, etc. To handle these challenges, an automatic Arabic text recognition requires building a robust system by computing discriminative features and applying a rigorous classifier together to achieve an improved performance. In this work, we introduce a new deep learning based system that recognizes Arabic text contained in images. We propose a novel hybrid network, combining a Bag-of-Feature (BoF) framework for feature extraction based on a deep Sparse Auto-Encoder (SAE), and Hidden Markov Models (HMMs), for sequence recognition. Our proposed system, termed BoF-deep SAE-HMM, is tested on four datasets, namely the printed Arabic line images Printed KHATT (P-KHATT), the benchmark printed word images Arabic Printed Text Image (APTI), the benchmark handwritten Arabic word images IFN/ENIT, and the benchmark handwritten digits images Modified National Institute of Standards and Technology (MNIST).

Handwritten character recognition (HCR) is presently gaining much popularity due to its wide range of applications. Bangla language comprises many unique characters of complex shapes, cursive nature, and resemblance among different characters. This nature of the script makes Bangla Handwritten character recognition (BHCR) a very challenging task. Very few HCR models can classify all types of Bangla characters with respectable accuracy. The current work proposes a hybrid BHCR model combining the Convolutional Neural Network (CNN) with stacked Bidirectional Long Short-Term Memory (Bi-LSTM). The experimental model used CNN for feature extraction and stacked Bidirectional LSTM for classification purposes. An augmented version of the CMATER dB dataset was implemented for training and evaluating the model. The model could recognize 243 individual classes of Bangla handwritten characters with 96.07% accuracy. No other proposed models are found to classify such a high number of classes with high accuracy.

Optical Character Recognition (OCR) is the process of extracting digitized text from images of scanned documents. While OCR systems have already matured in many languages, they still have shortcomings in cursive languages with overlapping letters such as the Arabic language. This paper proposes a complete Arabic OCR system that takes a scanned image of Arabic Naskh script as an input and generates a corresponding digital document. Our Arabic OCR system consists of the following modules: Pre-processing, Word-level Feature Extraction, Character Segmentation, Character Recognition, and Post-processing. This paper also proposes an improved font-independent character segmentation algorithm that outperforms the state-of-the-art segmentation algorithms. Lastly, the paper proposes a neural network model for the character recognition task. The system has experimented on several open Arabic corpora datasets with an average character segmentation accuracy 98.06%, character recognition accuracy 99.89%, and overall system accuracy 97.94% achieving outstanding results compared to the state-of-the-art Arabic OCR systems., 9 pages, 4 figures, 4 tables, 2 algorithms, accepted in the International Conference of Digital Image Processing and Pattern Recognition (DPPR), London, UK, November 2020

In this study, we approached Japanese cursive character (kuzushiji) recognition using machine learning to improve the efficiency of their transcription process. In particular, we discussed methods that automatically distinguish difficult-to-recognize characters or characters that were not included in the training, and put them on hold as unknown characters. We analyzed two methods that deferred recognition of characters: In the first method, characters with low confidence values obtained from the recognition model were put on hold as unknown characters. In the second method, we trained a “miscellaneous class” that combined multiple character classes and deferred the recognition of characters if they were recognized as belonging to the “miscellaneous class.” Experiments showed that by narrowing down the characters to be recognized, a recognition rate of more than 90% was achieved for over 1,500 character classes, confirming the usefulness of temporarily deferring recognition during the transcription process.

The recent increase in computational power and data available to the general public has given rise to the development of a plethora of highly performant deep neural network architectures. A popular and successful application of these has been in the field of image recognition, in which many yield promising results. Yet, there are differences and caveats in each, and knowing the right one to choose for a particular task can save a lot of money, time, and computational resources. In order to take a closer look at the differences in their performance, we have compared 5 of the many top architectures, namely the VGG16, VGG19, DenseNet121, ResNet50, and ResNet152V2 in terms of their error percentage, class-balanced error percentage, standard deviation of class-balanced accuracy, and training time. To achieve this, we have used 2 datasets, the Kuzushiji MNIST or KMNIST dataset, which is a balanced dataset with 10 classes and the Kuzushiji 49 or K49 dataset, an imbalanced dataset with 49 classes, both of which comprise cursive characters from the Japanese Hiragana script. All of the results were computed and verified using 10 fold cross validation. For the Kuzushiji MNIST dataset, the VGG16 showed the best results in all of the aspects. For the Kuzushiji 49 dataset, the DenseNet121 showed the highest performance in all aspects other than training time, and the latter was least for the VGG16.

Due to the nature of Arabic handwriting, segmenting words into characters/graphemes is the most difficult and critical task of the recognition system. The present paper proposes an approach to segment handwritten Arabic words into graphemes based on a directed Convolutional Neural Network (CNN) and Mathematical Morphology Operations (MMO). Arabic script is cursive, which means that almost all graphemes are connected via horizontal links; therefore, a technique to remove links will facilitate the segmentation of graphemes. In general, an MMO such as erosion seems suitable for getting the job done, but since Arabic handwriting is difficult, MMOs cause information loss and suffer from many issues such as diacritics and over-traces, which lead to over/under/bad segmentations. To overcome limitations, the present paper addresses these issues in the following order: the over-traces issue is addressed for the first time in the literature; a robust algorithm for diacritics extraction is provided; and finally, the main segmentation algorithm adopts a strategy based on a Partial Dilation (PD)-Global Erosion (GE) technique to combat the information loss issue. The PD phase amplifies important regions, while GE eliminates links between graphemes. The complementarity between PD and GE facilitates the extraction of graphemes and creates resistance against information loss. To properly tackle these difficult problems, this article exploits the robustness of CNNs, so a new directed CNN model is suggested. The idea is to draw the model's attention to certain targeted features, which are selected according to the nature of the problem addressed. The proposed directed CNN is used in all phases of the segmentation process. The experimental results are very encouraging and show that the proposed directed CNN model outperformed basic CNN in many experiments. The results also reveal that the followed strategy improved the ability of MMOs to perform segmentation and to compete with other approaches in this research area.

Handwritten character recognition is increasingly important in a variety of automation fields, for example, authentication of bank signatures, identification of ZIP codes on letter addresses, and forensic evidence. Despite improved object recognition technologies, Pashto&rsquo, s hand-written character recognition (PHCR) remains largely unsolved due to the presence of many enigmatic hand-written characters, enormously cursive Pashto characters, and lack of research attention. We propose a convolutional neural network (CNN) model for recognition of Pashto hand-written characters for the first time in an unrestricted environment. Firstly, a novel Pashto handwritten character data set, &ldquo, Poha&rdquo, for 44 characters is constructed. For preprocessing, deep fusion image processing techniques and noise reduction for text optimization are applied. A CNN model optimized in the number of convolutional layers and their parameters outperformed common deep models in terms of accuracy. Moreover, a set of benchmark popular CNN models applied to Poha is evaluated and compared with the proposed model. The obtained experimental results show that the proposed model is superior to other models with test accuracy of 99.64 percent for PHCR. The results indicate that our model may be a strong candidate for handwritten character recognition and automated PHCR applications.

In India, Devanagari script is used most prevalently. But recognizing the handwritten documents of Devanagari Script is a challenging task due to variations in writing, rendering, and cursive nature of handwriting. However, there has been a lot of improvement in research in the field of recognizing Handwritten Documents. In this, we will implement a framework that inputted the number of images gives an output for recognized characters.

Recognition of ancient Korean–Chinese cursive character (Hanja) is a challenging problem mainly because of large number of classes, damaged cursive characters, various hand-writing styles, and similar confusable characters. They also suffer from lack of training data and class imbalance issues. To address these problems, we propose a unified Regularized Low-shot Attention Transfer with Imbalance τ -Normalizing ( R E L A T I N ) framework. This handles the problem with instance-poor classes using a novel low-shot regularizer that encourages the norm of the weight vectors for classes with few samples to be aligned to those of many-shot classes. To overcome the class imbalance problem, we incorporate a decoupled classifier to rectify the decision boundaries via classifier weight-scaling into the proposed low-shot regularizer framework. To address the limited training data issue, the proposed framework performs Jensen–Shannon divergence based data augmentation and incorporate an attention module that aligns the most attentive features of the pretrained network to a target network. We verify the proposed RELATIN framework using highly-imbalanced ancient cursive handwritten character datasets. The results suggest that (i) the extreme class imbalance has a detrimental effect on classification performance; (ii) the proposed low-shot regularizer aligns the norm of the classifier in favor of classes with few samples; (iii) weight-scaling of decoupled classifier for addressing class imbalance appeared to be dominant in all the other baseline conditions; (iv) further addition of the attention module attempts to select more representative features maps from base pretrained model; (v) the proposed ( R E L A T I N ) framework results in superior representations to address extreme class imbalance issue.