Search

Handwritten Text Recognition has achieved an impressive performance in public benchmarks. However, due to the high inter- and intra-class variability between handwriting styles, such recognizers need to be trained using huge volumes of manually labeled training data. To alleviate this labor-consuming problem, synthetic data produced with TrueType fonts has been often used in the training loop to gain volume and augment the handwriting style variability. However, there is a significant style bias between synthetic and real data which hinders the improvement of recognition performance. To deal with such limitations, we propose a generative method for handwritten text-line images, which is conditioned on both visual appearance and textual content. Our method is able to produce long text-line samples with diverse handwriting styles. Once properly trained, our method can also be adapted to new target data by only accessing unlabeled text-line images to mimic handwritten styles and produce images with any textual content. Extensive experiments have been done on making use of the generated samples to boost Handwritten Text Recognition performance. Both qualitative and quantitative results demonstrate that the proposed approach outperforms the current state of the art., Comment: Accepted to TPAMI

The subtleties of human perception, as measured by vision scientists through the use of psychophysics, are important clues to the internal workings of visual recognition. For instance, measured reaction time can indicate whether a visual stimulus is easy for a subject to recognize, or whether it is hard. In this paper, we consider how to incorporate psychophysical measurements of visual perception into the loss function of a deep neural network being trained for a recognition task, under the assumption that such information can enforce consistency with human behavior. As a case study to assess the viability of this approach, we look at the problem of handwritten document transcription. While good progress has been made towards automatically transcribing modern handwriting, significant challenges remain in transcribing historical documents. Here we describe a general enhancement strategy, underpinned by the new loss formulation, which can be applied to the training regime of any deep learning-based document transcription system. Through experimentation, reliable performance improvement is demonstrated for the standard IAM and RIMES datasets for three different network architectures. Further, we go on to show feasibility for our approach on a new dataset of digitized Latin manuscripts, originally produced by scribes in the Cloister of St. Gall in the the 9th century.

Wouter Schouten (1638-1704) is the well-known writer of Oost-Indische voyagie, published in 1676, where he relates in vivid detail his experiences as a chirurgijn (medical assistant) in VOC service from 1658 till 1665, during which years his inquisitive nature brought him to the Moluccas and Celebes, and the Indian subcontinent, including Arakan (Burma) and Ceylon. A number of times he miraculously survived—shipwrecks and war with European competitors or indigenous rulers made life a hazardous enterprise. However, his rock-solid Calvinist faith never left him, and God’s blessing was, according to Schouten, responsible for his survival and the VOC successes. Schouten must have intended to add a poem to each chapter of the printed text of his itinerary. This plan came to nought, but two versions of the 36 poems are preserved in Schouten’s handwriting in the Koninklijke Bibliotheek in The Hague. In a fine edition these are now published for the first time. In his often extensive poems Schouten summarizes the events related in the prose version, adding comments and professing his devoutness. He certainly harboured literary pretensions, was well aware of the literary conventions of the time and the greater value ascribed to poetry in comparison to prose. After a brief general introduction the poems are included, with a summary

Numeral recognition plays a crucial role in creating automated systems such as posting address sorting and license plate recognition. Nowadays, numeral recognition systems which have the capability of recognising multiple languages are highly beneficial due to growing international correspondence and transactions, especially in multilingual countries where several languages are used simultaneously. Therefore, handwritten numeral recognition is more challenging than printed numeral recognition due to having different and complex handwriting styles. Hence, developing a multilingual handwritten system is considered as an important and debatable issue. We address this issue by proposing a language-independent model based on a robust CNN . Our proposed model is composed of language recognition and digit recognition, which aims to handle the recognition of multi-script images. We used transfer-learning in the proposed system to enhance the image quality and consequently the recognition performance. Extensive experiments were conducted to verify the effectiveness of both language and digit recognition procedures. The proposed system was tested with six different languages. The results showed an average accuracy of up to 99.8% for recognising various languages and the associated digits. The robustness and design procedure of the proposed model created a cost-effective extension for recognition of handwritten numerals in other languages.

Graphology is a method of assessment of human personality from individual's handwriting style. The personality in terms of various attributes like fear, honesty, work habits, social skills etc. can be profiled using Graphology. The personality profiling is based on various features like line spacing, margins, slant, letter size, text density etc. Letter t and i are very important personality trait indicators in graphology. Letter t is an indicator of will in individuals. Letter i is an indicator of social connectivity in individuals. This work proposes a deep learning classifier to classify the specific type of t and i patterns. Convolutional neural network is trained to extract the t and i features. The features extracted from t and i handwriting patterns are classified using the ensemble of Deep learning classifier. The performance of the proposed ensemble classification is tested against different handwritten documents. The proposed solution is able to achieve about 90% classification accuracy

Here we introduce Local Topological Recurrence Analysis (LoTRA), a simple computational approach for analyzing time-series data. Its versatility is elucidated using simulated data, Parkinsonian gait, and in vivo brain dynamics. We also show that this algorithm can be used to build a remarkably simple machine-learning model capable of outperforming deep-learning models in detecting Parkinson’s disease from a single digital handwriting test.

In handwriting recognition, traditional systems have relied heavily on handcrafted features and a massive amount of prior data and knowledge. Deep learning techniques have been the focus of research in the field of handwriting digit recognition and have achieved breakthrough performance in the last few years for knowledge extraction and management. KM and knowledge pyramid helps the project with its relationship with big data and IoT. The layers were selected randomly by which the performance of all the cases was found different. Data layers of the knowledge pyramid are formed by the sensors and input devices, whereas knowledge layers are the result of knowledge extraction applied on data layers. The knowledge pyramid and KM helps in making the use of IoT and big data easily. In this manuscript, the knowledge management principles capture the handwritten gestures numerically and get it recognized correctly by the software. The application of AI and DNN has increased the acceptability significantly. The accuracy is better than other available software on the market.

Arabic Handwritten Text Recognition (AHTR) based on deep learning approaches remains a challenging problem due to the inevitable domain shift like the variability among writers’ styles and the scarcity of labelled data. To alleviate such problems, we investigate in this paper different domain adaptation strategies of AHTR system. The main idea is to exploit the knowledge of a handwriting source domain and to transfer this knowledge to another domain where only few labelled data are available. Different writer-dependent and writer-independent domain adaptation strategies are explored using a convolutional neural networks (CNN) and Bidirectional Long Short Term Memory (BSTM) - connectionist temporal classification (CTC) architecture. To discuss the interest of the proposed techniques on the target domain, we have conducted extensive experiments using three Arabic handwritten text datasets, mainly, the MADCAT, the AHTID/MW and the IFN/ENIT. Concurrently, the Arabic handwritten text dataset KHATT was used as the source domain. The obtained results prove the effectiveness of the proposed strategies specially when considering the writer’s information during the supervised adaptation process.

Handwriting is used to distribute information among people. To access this information for further analysis the page needs to be optically scanned and converted to machine recognizable form. Due to unconstrained writing styles along with connected and overlapping characters, handwriting recognition remains a challenging task. Most of the methods in the literature use lexicon-based approaches and train their models on large datasets having near 50 K word samples to achieve good results. This results in high computational requirements. While these models use around 50 K words in their dictionary when recognizing handwritten English text, the actual number of words in the dictionary is much higher than this. To this end, we propose a handwriting recognition technique to recognize handwritten English text based on a YOLOv3 object recognition model that is lexicon-free and that performs sequential character detection and identification with a low number of training samples (only 1200 word images). This model works well without any dependency on writers’ style of writing. This is tested on the IAM dataset and it is able to achieve 29.21% Word Error Rate and 9.53% Character Error Rate without a predefined vocabulary, which is on par with the state-of-the-art lexicon-based word recognition models.

In this paper, we introduce a large scale dataset, called HKR, to address challenging detection and recognition problems of handwritten Russian and Kazakh text in the scanned documents. We present a new Russian and Kazakh database (with about 95% of Russian and 5% of Kazakh words/sentences respectively) for offline handwriting recognition. A few pre-processing and segmentation procedures have been developed together with the database. The database is written in Cyrillic and shares the same 33 characters. Besides these characters, the Kazakh alphabet also contains 9 additional specific characters. This dataset is a collection of forms. The sources of all the forms in the datasets were generated by LaTeXwhich subsequently was filled out by persons with their handwriting. The database consists of more than 1500 filled forms. There are approximately 63000 sentences, more than 715699 symbols produced by approximately 200 different writers. It can serve researchers in the field of handwriting recognition tasks by using deep and machine learning. For experiments, we used several popular text recognition methods for word and line recognition like CTC-based and attention-based methods. The results indicate the diversity of HKR. The dataset is available at https://github.com/abdoelsayed2016/HKR_Dataset .

At the present stage, the objects of forensic handwriting analysis are signatures performed with a significant gap in time. In this case, such objects are both signatures in studied documents, made at different intervals and studied signatures performed in relation to the comparative material with a time gap. Establishing stability of signature signs depending on performing time is possible taking into account characteristic signs of the stages of the signature handwriting formation. Since the factors causing changes in signatures are a lengthy process of forming the signature skill and the amount of signature practice (that is how often a person has to sign). While research on signatures that are at the formation stage, different signs, established together with significant coinciding signs canindicate a further improvement of movements. In personsof middle age (30-60 years old) who have a formed signature handwriting, significant changes do not occur over time until they stop practicing writing. While research on elderly people signatures along with the established coinciding signs, differing ones appear: signs of ataxia, decreased coordination of movements, low coherence of written signs, etc. These signs indicate degradation of movements due to physiological changes in the body in old age and a decrease in writing practice including practice of performing signature. Considering the above, research on signatures performed with a significant time gap is a rather complicated process that often causes difficulties when forensic expert evaluates revealed signs. Within the framework of this research paper, the authors focused on some topical issues existing in modern forensic expert practice of handwriting analyses of signatures performed with a significant time gap. The state of development of modern practical handwriting studies requires improvement of methodological approaches to this type of research. Currently, forensic experts are conducting R&D on the topic: “Improving research methods of signatures performed with a significant gap in time” which relevance is due to the need to improve forensic expert algorithm while handwriting research on signatures performed with a significant gap in time, as well as requirements of modern scientific standards.

In modern societies, handwritten identification is a great practical need and challenge for forensic sciences. Moreover, few research studies have focused on automatic offline handwritten document analysis. Also, there are very few standard datasets on handwritten document identification. In addition, handwritten documents lose their nature over time due to the spread and drying of ink. From this standpoint, the present study presents an offline writer identification system (in presence of the uncertainties such as different experimental conditions and environmental noises for more realistic assumptions) that is a critical need in forensic studies. For this purpose, a comprehensive right-to-left dataset is developed and devised by gathering data from 62 participants at different time intervals under different experimental conditions. This dataset is designed based on American Society for Testing and Materials (ASTM) standards. A Deep Convolutional Neural Network (DCNN) model based on modified pre-trained networks is designed and developed to extract features from raw data hierarchically. The proposed DCNN model is tested and investigated not only on the designed dataset but also on different datasets. One notable advantage of the present study is that it has made use of heterogeneous data. Another remarkable aspect of the proposed study is that the proposed DCNN model is independent of any specific languages which can be applied on various languages. The results of the study indicate that the proposed DCNN model can learn features hierarchically from the handwriting raw data and achieve higher accuracy than other comparative methods.

The recognition of handwritten Bangla digit is providing significant progress on optical character recognition (OCR). It is a very critical task due to the similar pattern and alignment of handwriting digits. With the progress of modern research on optical character recognition, it is reducing the complexity of the classification task by several methods, a few problems encounter during recognition and wait to be solved with simpler methods. The modern emerging field of artificial intelligence is the Deep Neural Network, which promises a solid solution to these few handwritten recognition problems. This paper proposed a fine regulated deep neural network (FRDNN) for the handwritten numeric character recognition problem that uses convolutional neural network (CNN) models with regularization parameters which makes the model generalized by preventing the overfitting. This paper applied Traditional Deep Neural Network (TDNN) and Fine regulated deep neural network (FRDNN) models with a similar layer experienced on BanglaLekha-Isolated databases and the classification accuracies for the two models were 96.25% and 96.99%, respectively over 100 epochs. The network performance of the FRDNN model on the BanglaLekha-Isolated digit dataset was more robust and accurate than the TDNN model and depend on experimentation. Our proposed method is obtained a good recognition accuracy compared with other existing available methods.

The lack of large training data in the context of deep learning applications is a serious issue investigated by many studies that deal with the current challenge. In this paper, we introduce new data augmentation methods that generate more shape and dynamic variations to improve the performance of recognition systems using small datasets. Four data augmentation strategies are employed in our work. The first strategy employs the geometric methods that include: italicity angle, change of magnitude ratio, and baseline inclination angle. The second strategy applies a frequency treatment that attenuates or amplifies the trajectory high harmonics to generate handwriting modified styles. The third strategy employs the beta-elliptic model to extract a combined static and dynamic representation of the handwritten trajectory which undergoes limited random change around its parameters in order to generate more modified samples. The hybrid strategy consists of combining these strategies to maximize variations of the online handwriting trajectory (OHT). We evaluated our approach of data augmentation in the context of multi-lingual online handwriting recognition (OHR) tasks using end-to-end CNN architecture. Four databases; ADAB, ALTEC-OnDB, and Online_KHATT for Arabic script, and UNIPEN for Latin characters, are used to validate the proposed strategy. The obtained results show the effectiveness and the advantage of the adopted strategies compared with those registered before database extension or reported in the state-of-the-art systems.

Illegible handwriting of doctors in prescriptions causing difficulty for pharmacists in providing exact medicine to patients. Every year about thousands of patients losing their lives throughout the world due to doctors sloppy handwriting skills. The solution for this problem is a voice prescription application that uses speech-to-text technology to make flawless and legible prescriptions. The generated prescription can be shared with the patient either in a soft copy or hard copy format. This application also maintains the medical records of patients for future purposes and also provides the facility to book an appointment with a doctor. The integrated AIML (Artificial Intelligence Markup Language) chatbot suggests a patient determine the type of doctor to be consulted based on symptoms. And also it provides a diagnosis for minute health issues.

The specially-abled face a lot of issues when it comes to taking up exams. They require specific assistance for taking up their exams known as a scribe. However in today’s world finding a scribe is of great difficulty. Even if the scribe is found there are a lot of other practical difficulties while using a scribe to write exams. One such practical difficulty in using a scribe to write exams is that the scribe may not be aware of the technical terms used by the candidate. This causes loss of marks for the candidate if the scribe makes an error. The scribe should also have a legible handwriting which again is a problem. Our project aims at eliminating the need for a scribe and allowing the candidate itself to take up their exams by leveraging the latest technologies such as speech recognition and synthesis. This is achieved with the help of an android application, using which the questions can be stored in a database using administrator login. A separate candidate login is given. The question will be read out to the candidate and the answer will be taken as a voice response.

The research explains and analyzes the most important digital methods that depend on graphic design of elements and icons. These help illiterate people comprehend shapes, words, letters, and numbers, which greatly affect the process of understanding the matrix and the information available through the applications which are used. These systems allow diversification and distribution of elements and spaces using colors and consider how the recipient perceives this information. Further, the research sheds light on the role of electronic games and other applications that have become an integral part of the information design process in a graphic form that follows modern rules and regulations in this field. The research presents foundations of designing educational applications, some samples of literacy applications and displaying samples of speech recognition and handwriting applications, the main requirements for designing an electronic application. Requirements. The research also identifies the problem in the form of questions, and the research continued the explanation to reach its goals, ending with several findings and recommendations.

Automatic pattern detection has become increasingly important for scholars in the humanities as the number of manuscripts that have been digitised has grown. Most of the state-of-the-art methods used for pattern detection depend on the availability of a large number of training samples, which are typically not available in the humanities as they involve tedious manual annotation by researchers (e.g. marking the location and size of words, drawings, seals and so on). This makes the applicability of such methods very limited within the field of manuscript research. We propose a learning-free approach based on a state-of-the-art Naïve Bayes Nearest-Neighbour classifier for the task of pattern detection in manuscript images. The method has already been successfully applied to an actual research question from South Asian studies about palm-leaf manuscripts. Furthermore, state-of-the-art results have been achieved on two extremely challenging datasets, namely the AMADI_LontarSet dataset of handwriting on palm leaves for word-spotting and the DocExplore dataset of medieval manuscripts for pattern detection. A performance analysis is provided as well in order to facilitate later comparisons by other researchers. Finally, an easy-to-use implementation of the proposed method is developed as a software tool and made freely available.

Handwriting Analysis is described as a scientific study for the analysis of handwriting. It is a way of interpreting and ability to learn from peculiarities in handwriting. Offline handwriting analysis is a traditional approach that cannot be used efficiently for analysis. Online handwriting analysis, on the other hand, can utilize various aspects like pressure on the pen, timestamp and other factors which help in improving the effectiveness of analysis. Learning disabilities are neurological processing problems which can hamper the learning of the children. Dysgraphia is a learning disability that mainly affects a child’s handwriting and motor skills. It is found in 10 to 30% of school-aged children. Dysgraphia can be diagnosed by therapists based on children’s handwriting samples and manual evaluation techniques. This method is lengthy and inaccurate. In this work, automatic identification methods for and classification of dysgraphia in children in the age group 7 to 12 is described. The method performs analyzing of the child’s writing dynamics via blueprints of the pressure the pen puts on paper with the pen’s movements and orientation with the use of a standardized digital writing pad and machine learning algorithms. It basically has two phases, the training phase, and testing phase. In the training phase, handwriting samples of known results are given to the system. Then the model is built using some classifier, Random forest or Support Vector Machine. Once the model is trained, then in testing phase this model is used for classification of unknown samples to predict whether the subject has dysgraphia or not. This is then used to check the accuracy of the designed system.

Handwritten text recognition has been an ongoing attractive task to research in the field of document analysis and recognition with applications in handwriting forensics, paleography, document examination, and handwriting recognition. In the present research, an automatic method of writer recognition is presented using digitized images of unconstrained texts. Despite the increasing efforts by prior literature on the different methods used for the same purpose, such methods performance, particularly their accuracy, has not been promising, leaving plenty of room for improvements. This method made use of codebook-based writer characterization, with each writing sample represented by a group of computed features from a primary and secondary codebook. The writings were then represented through the computation of the probability of codebook patterns occurrence, and the probability distribution was employed for each writer’s characterization. Writer identification process involved comparing two writings through the computation of the distances between their respective probability distribution. The study carried out experiments to determine the performance of the implemented method in light of rates of identification with the help of standard datasets, namely, KRDOH and IAM, the former being the most current and largest Kurdish handwritten datasets with 1076 writers, and the latter being a dataset containing 650 writers. The outcome of the experiments was promising with a rate of identification of 94.3%, with the proposed method outperforming the state-of-the-art methods by 2–3%.

Existing handwriting recognition solution on mobile app provides off premise service which means the handwriting is processed in overseas servers. Data sent to abroad servers are not under our control and could be possibly mishandled or misused. As recognizing handwriting is a complex problem, deep learning is needed. This research has the objective of developing an on premise Indonesian handwriting recognition using open source deep learning solution. Comparison of various deep learning solution to be used in the development are done. The deep learning solution will be used to build architectures. Various database format are also compared to decide which format is suitable to gather Indonesian handwriting database. The gathered Indonesian handwriting database and built architectures are used for experiments which consists of number of Convolutional Neural Network (CNN) layers, rotation and noise data augmentation, and Gated Recurrent Unit (GRU) vs Long Short Term Memory (LSTM). Experiment results shows that rotation data augmentation is the parameter to be change to improve word accuracy and Character Error Rate (CER). The improvement is 64.8% and 23.2% to 69.6% and 20.6% respectively.

Writer identification is the task of specifying the genuine writer according to their handwriting across a set of enrolled subjects which is a noteworthy research topic in the community of document analysis and recognition. In this paper, a novel framework based totally on identity vector is introduced for the online writer identification task. In the proposed framework, the sequence of extracted feature vectors from each handwriting sample is embedded into a fixed-length vector, referred to as identity vector (i-vector), to capture the long-term sequence-level writer-related characteristics, and then passed to the next stage for classification. Several techniques for feature normalization and intra-class variation reduction techniques in the i-vector domain such as within-class covariance normalization and regularized linear discriminant analysis are also investigated. We extensively evaluate the introduced framework on the popular database, CAISA, for English and Chinese language in various scenarios, such as multi-language and cross-language. Experimental results show, in the best cases, the proposed framework could achieve 98.68% accuracy on English dataset and 96.03% on Chinese dataset of the CAISA database. These obtained results indicate an improvement over the best reported result of the current state-of-the-art approaches with the exception of fully end-to-end approaches which have their own serious limitation in the real applications. In addition to the accuracy improvement, due to its low computational load it has the potential to be implemented on the handheld digital devices.

This paper introduces a large-scale Meitei Mayek handwritten character database. It consists of the complete character set of the script. There are a total of 85,124 character images of 55 character classes with 72,330 and 12,794 images in training and test sets, respectively. The present work focuses on collecting the natural handwriting of individuals by carrying out sample collection in two phases: (a) unconstrained handwriting in the form of answer sheets and classroom notes and (b) tabular forms. A total of nearly 500 individuals have contributed in the development of the database. Recognition of the character images in the database is carried out using different feature descriptors with four popular classifiers, namely KNN, Linear Support Vector Classifier, Random Forest and Support Vector Machine. The paper also proposes a convolutional neural network (CNN) model by enhancing a base CNN architecture by optimally tuning the hyperparameters. Experimental results show that the CNN model can be benchmarked against the concerned database with a test accuracy of 95.56%.

Do handwriting skills transfer when a child writes in two different scripts, such as the Latin and Cyrillic alphabets? Are our measures of handwriting skills intrinsically bound to one alphabet or will a child who faces handwriting difficulties in one script experience similar difficulties in the other script? To answer these questions, 190 children from grades 1–4 were asked to copy a short text using both the Cyrillic and Latin alphabets on a digital tablet. A recent change of policy in Kazakhstan gave us an opportunity to measure transfer, as the Latin-based Kazakh alphabet has not yet been introduced. Therefore, pupils in grade 1 had a 6-months experience in Cyrillic, and pupils in grades 2, 3, and 4 had 1.5, 2.5, and 3.5 years of experience in Cyrillic, respectively. This unique situation created a quasi-experimental situation that allowed us to measure the influence of the number of years spent practicing Cyrillic on the quality of handwriting in the Latin alphabet. The results showed that some of the differences between the two scripts were constant across all grades. These differences thus reflect the intrinsic differences in the handwriting dynamics between the two alphabets. For instance, several features related to the pen pressure on the tablet are quite different. Other features, however, revealed decreasing differences between the two scripts across grades. While we found that the quality of Cyrillic writing increased from grades 1–4, due to increased practice, we also found that the quality of the Latin writing increased as well, despite the fact that all of the pupils had the same absence of experience in writing in Latin. We can therefore interpret this improvement in Latin script as an indicator of the transfer of fine motor control skills from Cyrillic to Latin. This result is especially surprising given that one could instead hypothesize a negative transfer, i.e., that the finger controls automated for one alphabet would interfere with those required by the other alphabet. One interesting side-effect of these findings is that the algorithms that we developed for the diagnosis of handwriting difficulties among French-speaking children could be relevant for other alphabets, paving the way for the creation of a cross-lingual model for the detection of handwriting difficulties.

The development of pattern recognition and artificial intelligence domains owes the writer identification challenge greatly. In fact, writer identification is still a challenging task in the definition of a set of features able to characterize the various handwritten documents. These handwritten documents are not generally stable and show a wide variability from the same person over time, or from different writers. The capacity to identify the documents’ writers provides further chances of using these handwritten documents for several applications like forensic science, control access, digital rights management and financial transactions. In this paper, we propose a novel system to text-independent online multilingual writer identification. Our system is based on new model that we named the Extended Beta-Elliptic Model. Moreover, we are interested in using the Fuzzy Elementary Perceptual Codes to characterize the handwriting of writers well. In addition, we adopted the use of Recurrent Neural Network with Deep Bidirectional Long Short-Term Memory in the training and identification phases. Experiments are conducted on IBM_UB_1 and ADAB datasets with 98.44% and 100% writer identification rates respectively. The proposed system using the combination of the Extended Beta-Elliptic model and the Fuzzy Elementary Perceptual Codes in features extraction and the Deep Bidirectional Long Short-Term Memory in classification outperforms the existing online writer identification systems on both Latin and Arabic scripts.

Electronic exam's' non-reliability is one of the most critical gaps in this type of exam, whereas any electronic exam is considered entirely unreliable. In addition to that, it is not protected and free from cheat. The continuous validation that the examinee is the one who solves the questions of the electronic exam permanently and continuously is the most crucial goal of this research. Many methods and procedures will be discussed, which will protect the validity of this type of exam. For example, one of these methods matches the student's fingerprint and takes random snapshots while taking the electronic exam in more than one format. In this research, it's proposed to use student handwriting during the electronic exams to built the value of validity. This way is considered one of the rare methods in the electronic exam. The research also tackles the process of reality for electronic exams through a fingerprint during the examination period. It takes random snapshots for the student without paying attention to him or appearing those pictures to the observer or correcting the exam. The difficulty of offering a safe and high-credibility exam and reducing the phenomenon of cheating requires excellent capabilities in some countries. It also needs prior arrangements and good preparation. It also needs to develop technological capabilities continuously in a high-cost material that may reduce the cheating rates in such a kind of exam.

Demographical handwritings classification has many applications in various disciplines such as biometrics forensics, psychology, archeology, etc. Finding the best features for differentiating subclasses (e.g. men and women) is one of the major problems in handwriting based demographical classification. Convolutional Neural Networks (CNNs) advanced models have a higher capacity in extracting appropriate features compared to traditional models. In this paper, the ability and capacity of deep CNNs in automatic classification of two handwriting based demographical problems, i.e. gender and handedness classification, have been examined by using advanced CNNs; DenseNet201, InceptionV3, and Xception. Two databases, IAM (English texts) and KHATT (Arabic texts) have been employed in this study. The achieved results showed that the proposed CNNs architectures performed well in improving classification results, with 84% accuracy (1.27% improvement) for gender classification using the IAM database, and 99.14% accuracy (28.23% improvement) for handedness classification using the KHATT database.

The purpose of the article is to study the issues of qualification of objects of forensic handwriting examination. Handwritten notes made on different surfaces in the system of objects of handwriting examination are considered. Specific features of the appointment and performance of handwriting examinations in relation to handwriting objects made on different surfaces are presented. It is concluded that the identification of the performer of a particular recording is possible only when the features of the performer’s handwriting are sufficiently reflected, which in its totality are individual, that is, inherent only in the handwriting of a certain person. As a result, individual strokes and numbers that do not reflect the totality of signs of the writing of their performers, as a rule, do not allow establishing the artist of these recordings. Despite a number of factors that make it difficult to study handwritten objects made on various surfaces, the solution of identification and some diagnostic problems for this type of objects is still possible. This possibility is determined not only by the output characteristics of the handwritten note being studied, but also by the preparation for conducting the corresponding research at all stages of the appointment and carrying out forensic handwriting examination.

Writer identification is the procedure of identifying individuals from handwriting. Writer identification is a common interest in biometric authentication and verification systems, and numerous studies are available for English, Chinese, Arabic, and Persian specific handwriting. This paper introduces a supervised offline Indic script writer identification system that can identify individuals using less than a single page of handwriting. A lightweight Convolutional Neural Network (CNN) architecture fused with non-trainable Gabor filters is used as an identification model that can recognize writers from scarce data. For the experiment, we used BanglaWriting dataset, which is openly available for Bengali writing and writer recognition. Further, we added Devanagari and Telugu datasets for evaluation. The overall evaluation shows that the proposed thresholded Gabor-based CNN architecture performs superior to numerous deep CNN architectures for Indic writer recognition.

The last decades have been characterized by the rapid development of computer technologies and its widespread use in various fields of science, education, production, etc. The sphere of law enforcement is no exception, including the use of these technologies for conducting expert research. At the same time, certain problems of digitalization (translation of information into digital form) of technical and forensic studies and, in particular, studies of handwriting studies, are still observed. The article deals with the issue of the possibility of using the computer system AutoCAD in determining the identification signs in the handwriting examination of signatures. Comparing the numerical values of handwriting features obtained for a number of samples and for the objects under study, at the stage of comparative research, it is possible to establish whether the resulting numerical values characterizing some individual graphic features of the signature under study do not go beyond the variability limits determined for samples of signatures of a certain person Perhaps, by deepening and developing this method, it will be possible to identify such parameters that are characteristic only of the signature of a certain person. It should be noted that the proposed method of numerical expression of some graphic features of signatures does not cover all the variety of these features. These are signs related to the shape of the movements, direction, curvature, sequence of execution, the method of beginning and ending movements, etc. Such incompleteness needs to be carefully refined and opens wide opportunities for further research.

Segmentation of a touching component to separate its constituent text and nontext parts is always a very crucial but challenging task toward developing a comprehensive document image processing (DIP) system. This is because, irrespective of document types, either printed or handwritten, the nontext parts need to be suppressed first before processing the text parts through an optical character recognition (OCR) system. Although a good number of attempts have been made to address this issue for printed documents, the same for regular handwritten document images is almost none. However, the appearance of touching components where a nontext part gets joined with a text part is a common issue in freestyle handwriting. To this end, in this work, we tailor-make a generative adversarial network (GAN)-based model with a suitable loss function that we name tsegGAN. We also prepare an in-house data set by collecting touching components from different real-world handwritten documents to evaluate our model. The performance comparison of our model with state-of-the-art GAN models shows that tsegGAN has outperformed the others with a significant margin.

Automatically generating handwritten text images is a challenging task due to the diverse handwriting styles and the irregular writing in natural scenes. In this paper, we propose an effective generative model called HTG-GAN to synthesize handwritten text images from latent prior. Unlike single-character synthesis, our method is capable of generating images of sequence characters with arbitrary length, which pays more attention to the structural relationship between characters. We model the structural relationship as the style representation to avoid explicitly modeling the stroke layout. Specifically, the text image is disentangled into style representation and content representation, where the style representation is mapped into Gaussian distribution and the content representation is embedded using character index. In this way, our model can generate new handwritten text images with specified contents and various styles to perform data augmentation, thereby boosting handwritten text recognition (HTR). Experimental results show that our method achieves state-of-the-art performance in handwritten text generation.

The thesis is aimed at considering situations that arise during the performance of handwriting examinations in which conclusions are possible that have various legal consequences for the parties to the trial. This is because of the fact that when assigning a handwriting examination, there are situations when a small number of signature samples are provided for a comparative study or the interested party of the torturer misleads the expert by providing false signature samples. By carefully studying the case materials, the expert manages to establish additional samples of signatures, which are not mentioned in the document on the appointment of a handwriting examination. It helps the expert to determine and come to a categorical conclusion which case the signature in the document under study was affixed not by the person himself, but by another person. Or confirm the fact with one hundred percent certainty that the signature was made by one of the relatives. In such cases, if it were not for additional samples from the case materials, it is possible for an expert to come to a false conclusion. The question may arise whether the expert violates sub-clause 3, clause 2.3. Instructions, the expert is prohibited from: “independently collecting materials that are subject to research, as well as choosing outgoing data for the examination, if they are displayed ambiguously in the materials provided to him/her”. But clause 2.1. of the Instruction state that the expert has the right to indicate in his conclusion the facts revealed during the examination that are significant for the case, but regarding which he/she was not asked questions and the circumstances that contributed (could contribute) to the commission of the offense.

The modern capabilities of computers have returned interest in artificial intelligence technologies. A particular area of application of these technologies is pattern recognition, which can be applied to the traditional forensic task – identification of signs of forgery (imitation) of a signature. The results of forgery are differentiated into three types: auto-forgery, simple and skilled forgeries. Only skilled forgeries are considered in this study. The online and offline approaches to the study of signatures and other handwriting material are described. The developed artificial intelligence system based on an artificial neural network refers to the offline type of signature recognition – that is, it is focused on working exclusively with the consequences of the signature – its graphic image. The content and principles of the formation of a hypothesis for the development of an artificial intelligence system are described with a combination of humanitarian (legal) knowledge and natural-technical knowledge. At the initial stage of the study, in order to develop an experimental-applied artificial intelligence system based on an artificial neural network focused on identifying forged signatures, 127 people were questioned in order to identify a person's ability to detect fake signatures. It was found that under experimental conditions the probability of a correct determination of the originality or forgery of the presented signature for the respondent is on average 69.29 %. Accordingly, this value can be used as a threshold for determining the effectiveness of the developed artificial intelligence system. In the process of preparing the dataset (an array for training and verification of its results) of the system in terms of fraudulent signatures, some forensically significant features were revealed, associated with the psychological and anatomical features of the person performing the forgery, both known to criminalistics and new ones. It is emphasized that the joint development of artificial intelligence systems by the methods of computer science and criminalistics can generate additional results that may be useful outside the scope of the research tasks.

Riau province has Malay Arabic script as a traditional cultural heritage of ancient characters that should be preserved; this script is adapted from Arabic writing. This script from Malay Arabic has a unique form that is different from the original Arabic writing adaptation, which is read in a combination of letters forming latin meanings as an introduction to the everyday language of Riau Malay people in the earlier kingdom. Malay Arabic writing became an introduction to the local content of traditional languages in schools. To foster a love for preserving culture, in accordance with current technology that is able to recognize scripting patterns when written in paper, a knowledge base was created by using Matlab software by applying a convolutional Neural Network (CNN) artificial neural network algorithm capable of recognizing script patterns well. The result of image input in the form of handwriting written on paper then in the scanner in the form of JPEG image format. Testing was carried out on four Arabic Malay characters namely alif, ha, la, kho and nun. The result of training for the letter alif (a) epoch is obtained 98 out of 100 iterations with a training length of 3 seconds, furthermore, in validation performance with a result of 0.25013 on epoch 92 of 98 epoch for gradient letters with a value of 0.0071991 on the next epoch 98 in the extras produces an accuracy value of 0.6548 which states the correct result accordingness because it is close to the alif script. In the process of train input the letter kho obtained epoch 80 out of 100 iterations with a training process for 3 seconds, validation performance 0.25153 on epoch 74 out of 80 epoch for check validation with a value of 0.0011682 on the next epoch 80 in the extras obtained an extra value of 0.9326 stated the value is incorrect. Because the result of the extras results in an image that does not come close to the kho letter. Therefore, a study of how the system can recognize Malay Arabic writing patterns with the Convolutional Neural Network (CNN) method because it is very good at identifying image pattern features with an accuracy value of 4.12% of the 10 sample image patterns that have been inputted. With the introduction of imagery patterns from the extraction of features scanned Malay Arabic characters can help the findings of ancient Malay Arabic script as morphological learning of the validity of abstraction of Malay Arabic script is good

Optical Character Recognition (OCR) system is used to convert the document images, either printed or handwritten, into its electronic counterpart. But dealing with handwritten texts is much more challenging than printed ones due to the erratic writing style of the individuals. The problem becomes more severe when the input image is a doctor’s prescription. Before feeding such an image to the OCR engine, the classification of printed and handwritten texts is a necessity as a doctor’s prescription contains both handwritten and printed texts which are to be processed separately. Much work has been done in the domain of handwritten and printed text separation albeit work related to doctor’s handwriting. In this paper, a method is proposed which first localizes the position of texts in a doctor’s prescription, and then separates out the printed texts from the handwritten ones. Due to the unavailability of a large database, we have used some standard data (image) augmentation techniques to evaluate as well as to prove the robustness of our method. Besides, we have also designed a Graphical User Interface (GUI) so that anybody can visualize the output by providing a prescription image as input.

Today, people rely almost full time on digital texts. It is not surprising that handwriting earned a special status, and solutions to mimic real handwriting became attractive. A particular field called handwriting synthesis generates renderings of text which resemble natural writing but are synthesized from actual handwriting samples. The main idea behind samples’ current solutions is to collect enough samples to capture a given subject’s writing style, and therefore be able to reproduce it in new texts, with natural variability. Nevertheless, the question remains of how much input variability is enough to represent specific handwriting. In this paper, we address sample acquisition for handwriting synthesis. We conducted a study comparing written text similarity between two sets of samples, one using augmented pangrams (with a total of 473 characters) and the other using general texts (with 1586 characters). Our results show that the samples collected with pangrams are statistically equivalent in variation with samples collected using general texts, with many benefits, particularly the shorter time needed to collect the samples. We also made our data collection publicly available, providing a valuable original resource for future research.

Статья посвящена атрибуции почерков рукописей, посвящённых жизнеописанию афонского схимонаха Селевкия. Выявлены разные почерки, отмечены текстологические и лингвистические особенности рукописей, находящихся в книгохранилищах Афона и Москвы, считающиеся его автографом, но таковыми на самом деле не являющиеся. This article deals with the attribution of handwritings of manuscripts dedicated to the biography of the Athonite schema Seleucus. Different handwritings are identified and the textological and linguistic features of manuscripts in the book repositories of Mount Athos and Moscow, which are believed to be his autograph, but are not in fact so, are noted.

Introduction: In Taiwan, although many school-aged students with low vision struggle with poor handwriting, there is a lack of evidence-based educational practices for handwriting enhancement. Since the use of mobile technology has increased recently, iPads have been identified as an effective tool to deliver video-based instruction to individuals. In this study, iPads were used to deliver video modeling to provide instruction designed to enhance the handwriting accuracy among students with low vision. Methods: A multiple-baseline-across-participants-probe design was used in this study to assess the success of the intervention with three individuals with low vision who were 9, 12, and 14 years of age. Results: During the baseline phase, the accuracy for each of the participants was under 50%. After intervention, their handwriting accuracy increased. The participants were also able to maintain their performance during follow-up sessions. Moreover, the study demonstrated good social validity, since the stakeholders all indicated a high level of treatment acceptability for this intervention. Discussion: The outcome of this study demonstrates that iPads used with video modeling can lead to improved accuracy in handwriting for students with low vision. The findings also support video modeling as an effective strategy for teaching new skills and may be applied to students with different special learning needs. Implications for practitioners: Video modeling is an effective and feasible instructional strategy for practitioners as it can be easily implemented. Additionally, given its built-in visual support, the iPad is an effective instrument that can help students with low vision reach a higher potential for handwriting accuracy. It is worth noting that, in addition to the presence of video modeling, proper instruction, as well as opportunities to practice, is needed for students to produce accurate word formation. Thus, short daily practices combined with the intervention method presented in this study are likely to gain better results for students with low vision.

Pengembangan perangkat lunak yang baik harus melalui proses pengujian tiap fase. Pengujian digunakan untuk mencari kesalahan perangkat lunak yang dikembangkan. Pengujian dilakukan berdasarkan kebutuhan pengembangan perangkat lunak. Pengujian ini bertujuan memvalidasi kondisi awal dan kondisi akhir sesuai dengan kebutuhan. Pada penelitian ini pengujian antarmuka dilakukan dengan model behaviour Unifed Modelling Language (UML) yaitu use case diagram. Berdasarkan hasil uji, terdapat 15 pengujian yang dilakukan sesuai use case scenario. Terdapat 1 pengujian yang ditolak, hal ini dilakukuan perbaikan pada fungsionalitas antarmuka aplikasi pengenalan tulisan tangan ini. Sehingga hasil pengujian menunjukan bahwa aplikasi dapat berjalan sesuai scenario. Bedasarkan pengujian ini dapat mengetahui kualitas dari aplikasi pengenelan tulisan tangan.

The possibility of refi ning and unifying the algorithm of preliminary examination of copies of documents to establish their suitability/unsuitability for further research is considered. Examples of the possibility of establishing the fact of signs of editing in copies of documents presence are given. Empowerment the fi eld of handwriting examiners training through introducing special knowledge from the fi elds of questioned documents examination and computer forensic into the training program is proposed. The stages of the proposed algorithm are analyzed sequentially. The author’s algorithm for preliminary examination of copies of documents containing handwritten requisites images is given. Arguments in favor of the practical possibility of producing handwriting examination of copies of handwritten requisites are given. The correlation of examination of copies and originals of handwritten requisites have been undertaken. Illustrations demonstrating the dependence of the ability to evaluate some features on the resolution of the image of a handwritten object are presented.

Handwriting recognition remains a challenge in the machine vision field, especially in optical character recognition (OCR). The OCR has various applications such as the detection of handwritten Farsi digits and the diagnosis of biomedical science. In expanding and improving quality of the subject, this research focus on the recognition of Farsi Handwriting Digits and illustration applications in biomedical science. The detection of handwritten Farsi digits is being widely used in most contexts involving the collection of generic digital numerical information, such as reading checks or digits of postcodes. Selecting an appropriate classifier has become an issue highlighted in the recognition of handwritten digits. The paper aims at identifying handwritten Farsi digits written with different handwritten styles. Digits are classified using several traditional methods, including K-nearest neighbor, artificial neural network (ANN), and support vector machine (SVM) classifiers. New features of digits, namely, geometric and correlation-based features, have demonstrated to achieve better recognition performance. A noble class of methods, known as deep neural networks (DNNs), is also used to identify handwritten digits through machine vision. Here, two types of introduce its expansion form, a convolutional neural network (CNN) and an auto-encoder, are implemented. Moreover, by using a new combination of CNN layers one can obtain improved results in classifying Farsi digits. The performances of the DNN-based and traditional classifiers are compared to investigate the improvements in accuracy and calculation time. The SVM shows the best results among the traditional classifiers, whereas the CNN achieves the best results among the investigated techniques. The ANN offers better execution time than the SVM, but its accuracy is lower. The best accuracy among the traditional classifiers based on all investigated features is 99.3% accuracy obtained by the SVM, and the CNN achieves the best overall accuracy of 99.45%.

When filling out privacy-related forms in public places such as hospitals or clinics, people usually are not aware that the sound of their handwriting leaks personal information. In this paper, we explore the possibility of eavesdropping on handwriting via nearby mobile devices based on audio signal processing and machine learning. By presenting a proof-of-concept system, WritingHacker, we show the usage of mobile devices to collect the sound of victims’ handwriting, and to extract handwriting-specific features for machine learning based analysis. WritingHacker focuses on the situation where the victim's handwriting follows certain print style. An attacker can keep a mobile device, such as a common smartphone, touching the desk used by the victim to record the audio signals of handwriting. Then, the system can provide a word-level estimate for the content of the handwriting. To reduce the impacts of various writing habits and writing locations, the system utilizes the methods of letter clustering, dictionary filtering and letter time length based offsetting. Moreover, if the relative position between the device and the handwriting is known, a hand motion tracking method can be further applied to enhance the system's performance. Our prototype system's experimental results show that the accuracy of word recognition reaches around 70 - 80 percent under certain conditions, which reveals the danger of privacy leakage through the sound of handwriting.

Automatic handwriting recognition is an important component for many applications in various fields. It is a challenging problem that has received a lot of attention in the past three decades. Research has focused on the recognition of Latin languages’ handwriting. Fewer studies have been done for the Arabic language. In this paper, we present a new dataset of Arabic letters written exclusively by children aged 7–12 which we call Hijja. Our dataset contains 47,434 characters written by 591 participants. In addition, we propose an automatic handwriting recognition model based on convolutional neural networks (CNN). We train our model on Hijja, as well as the Arabic Handwritten Character Dataset (AHCD) dataset. Results show that our model’s performance is promising, achieving accuracies of 97% and 88% on the AHCD dataset and the Hijja dataset, respectively, outperforming other models in the literature.

Artificial intelligence methods are used in many fields of applications. One of them is biometrics. It deals with a broadly understood analysis of physical and behavioral biometric characteristics. Behavioral characteristics are particularly important from a practical point of view. They describe, among others, learned behaviors of human beings. They include the ways of signing, lip movement, gait, etc. Such features are difficult to forge, so their use in biometric systems increases security. The dynamics of signing is particularly important because it is considered socially acceptable worldwide. However, the problem associated with its processing results from its possible instability over time, as the signature is likely to evolve. In a situation where the method of signing is only partially altered (and not completely changed), the trend in possible changes can be predicted. In this article, we propose a new approach to predicting such changes. For this purpose, we use the possibilities offered by fuzzy systems (FSs) and population-based algorithms (PBAs). The purpose of applying the PBA is to select parameters and the structure of the FSs and to select reference signatures for each user separately so as to maximize prediction accuracy. Prediction results are included in the final verification of signatures. An important feature of the prediction mechanism is also its ability to interpret the changes occurring in signature handwriting because they are represented by the fuzzy rules base. This type of a flexible and nature-inspired prediction mechanism would be very difficult to implement with the use of other methods. The developed mechanisms can be used in various fields of applications—also in those not related to biometrics.

The article is devoted to the consideration of working situations that may arise with comparative samples of signatures provided for research during the relevant examinations. Possible changes in regulatory legal acts that will optimize the work of an expert during judicial handwriting examinations are highlighted. The main goal of this article is to consider the practical situations that an expert has when studying and working with comparative samples received for research. An analysis of expert practice will help minimize expert errors, which in turn will affect the objectivity of an expert’s opinion. The article sets out the situations that arise at the stage of studying materials that can affect the objectivity and categorization of an expert’s opinion: -providing free samples of signatures in documents issued to another person; -providing free samples of signatures in documents with amendments, supplements, incorrect indication of the name, and signature of the current user of the subscriber book or the owner of the document (passport). The situations discussed will help interns, novice experts in the study of materials received for research. It should be noted that a superficial approach at the stage of their research might lead to the need for additional examinations. The article can be useful to judges, assistant judges, representatives of law enforcement agencies, lawyers when assigning a handwriting examination and can help to reduce the time for the examination.