Handwritten Character Images Research Articles

Incomplete handwritten Dongba character image recognition by multiscale feature restoration

Incomplete handwritten Dongba character often appears in heritage documents and its recognition is significant for heritage and philology. However, all previous methods always suppose that a complete Dongba character is used as input, and thus fail to achieve satisfactory performance when applied to incomplete Dongba character recognition. In this paper, an end-to-end network (DB2RNet) is proposed for incomplete handwritten Dongba character image recognition by multiscale feature restoration. Specifically, we first develop datasets that contain different levels of incomplete Dongba characters. A restoration module is proposed to restore the input incomplete Dongba character, and then a recognition module is employed to recognize Dongba character. By introducing an inter-module residual connection between the restoration module and recognition module, the DB2RNet can strengthen feature information transmission and boost the recognition performance. In addition, novel multiscale feature blocks are introduced, which can provide more effective texture and contextual information transmission for Dongba character image restoration, and thus yielding better restoration effects and better recognition results. Extensive experiments are conducted on Dongba character, Chinese character and Oracle character datasets and validate the effectiveness, superiority and robustness of our methods. Experiments results demonstrate that our proposed DB2RNet achieves competitive Dongba character restoration and recognition performance and outperforms the current state-of-the-art methods.

On-device Online Learning and Semantic Management of TinyML Systems

Recent advances in Tiny Machine Learning (TinyML) empower low-footprint embedded devices for real-time on-device Machine Learning (ML). While many acknowledge the potential benefits of TinyML, its practical implementation presents unique challenges. This study aims to bridge the gap between prototyping single TinyML models and developing reliable TinyML systems in production: (1) Embedded devices operate in dynamically changing conditions. Existing TinyML solutions primarily focus on inference, with models trained offline on powerful machines and deployed as static objects. However, static models may underperform in the real world due to evolving input data distributions. We propose online learning to enable training on constrained devices, adapting local models toward the latest field conditions. (2) Nevertheless, current on-device learning methods struggle with heterogeneous deployment conditions and the scarcity of labeled data when applied across numerous devices. We introduce federated meta-learning incorporating online learning to enhance model generalization, facilitating rapid learning. This approach ensures optimal performance among distributed devices by knowledge sharing. (3) Moreover, TinyML’s pivotal advantage is widespread adoption. Embedded devices and TinyML models prioritize extreme efficiency, leading to diverse characteristics ranging from memory and sensors to model architectures. Given their diversity and non-standardized representations, managing these resources becomes challenging as TinyML systems scale up. We present semantic management for the joint management of models and devices at scale. We demonstrate our methods through a basic regression example and then assess them in three real-world TinyML applications: handwritten character image classification, keyword audio classification, and smart building presence detection. The results confirm the effectiveness of our approaches from various perspectives, such as accuracy improvement, resource savings, and engineering effort reduction.

A study on offline handwritten Chinese character recognition based on chaotic iteration

Offline handwritten Chinese character recognition technology has important practical value in the field of automatic machine entryomit, and is one of the research hotspots of intelligent technology. In this paper, a handwritten Chinese character feature extraction method based on dynamic system iteration is constructed by using chaos theory. Firstly, from the perspective of system iteration, the Chinese character image is preprocessed to optimize the image information. Secondly, the iterative matrix constructed by the auxiliary function is combined with the handwritten Chinese character image dot matrix to avoid repetiton the discrete dynamic system, and the approximate chaotic attractor is iteratively generated as the image feature. Then, the similarity of the image is determined according to the size of the Pearson correlation coefficient. Finally, multiple auxiliary functions with large differences are selected to generate multiple discrete chaotic dynamic systems. The above operations are repeated, and the unrecognized image set is used as the next data set to be identified. The experimental results show that the recognition algorithm has achieved good recognition results on the HCL2000 Chinese character database.

Recognition of Off-Line Handwritten Kannada Words Using Enhanced Skew Detection and Correction Method

Handwritten character image is taken as dataset for this method. In the proposed system, extracting text lines, word, charactersand skew correction are done based on Enhanced Skew Detection and Correction for Words algorithm for estimating and correcting skew lines. The algorithm will be used for finding the height andwidth of the entire handwritten word. In case of no skew, minimum valuewill be considered for the height of the wordand maximum value will be considered for the width of the word. Once skew is corrected with approximateskew angle repetition of the same process, only busy zone is considered for performing precise skewcorrection. The recommended approach has tested roughly 3364 Kannada items and achieved the best performance of 97.05 percent.

Генерация изображений рукописного текста на русском языке

Automatic handwriting recognition is an important component in the process of electronic documents analysis, but its solution is still far from ideal. One of the main reasons for the complexity of Russian handwriting recognition is the insufficient amount of data used to train recognition models. Moreover, for the Russian language the problem is more acute and is exacerbated by a large variety of complex handwriting. This paper explores the impact of various methods of generating additional training datasets on the quality of recognition models: the method based on handwritten fonts, the StackMix method of gluing words from symbols, and the use of a generative adversarial network. A font-based method for creating images of handwritten text in Russian has been developed and described in this work. In addition, an algorithm for the formation of a new Cyrillic handwritten font based on the existing images of handwritten characters is proposed. The effectiveness of the developed method was tested using experiments that were carried out on two publicly available Cyrillic datasets using two different recognition models. The results of the experiments showed that the developed method for generating images made it possible to increase the accuracy of handwriting recognition by an average of 6%, which is comparable to the results of other more complex methods. The source code of the experiments, the proposed method, as well as the datasets generated during the experiments are posted in the public domain and are ready for download.

DeblurGAN-CNN: Effective Image Denoising and Recognition for Noisy Handwritten Characters

Many problems can reduce handwritten character recognition performance, such as image degradation, light conditions, low-resolution images, and even the quality of the capture devices. However, in this research, we have focused on the noise in the character images that could decrease the accuracy of handwritten character recognition. Many types of noise penalties influence the recognition performance, for example, low resolution, Gaussian noise, low contrast, and blur. First, this research proposes a method that learns from the noisy handwritten character images and synthesizes clean character images using the robust deblur generative adversarial network (DeblurGAN). Second, we combine the DeblurGAN architecture with a convolutional neural network (CNN), called DeblurGAN-CNN. Subsequently, two state-of-the-art CNN architectures are combined with DeblurGAN, namely DeblurGAN-DenseNet121 and DeblurGAN-MobileNetV2, to address many noise problems and enhance the recognition performance of the handwritten character images. Finally, the DeblurGAN-CNN could transform the noisy characters to the new clean characters and recognize clean characters simultaneously. We have evaluated and compared the experimental results of the proposed DeblurGAN-CNN architectures with the existing methods on four handwritten character datasets: n-THI-C68, n-MNIST, THI-C68, and THCC-67. For the n-THI-C68 dataset, the DeblurGAN-CNN achieved above 98% and outperformed the other existing methods. For the n-MNIST, the proposed DeblurGAN-CNN achieved an accuracy of 97.59% when the AWGN+Contrast noise method was applied to the handwritten digits. We have evaluated the DeblurGAN-CNN on the THCC-67 dataset. The result showed that the proposed DeblurGAN-CNN achieved an accuracy of 80.68%, which is significantly higher than the existing method, approximately 10%.

Convolutional Neural Network Based Intelligent Handwritten Document Recognition

This paper presents a handwritten document recognition system based on the convolutional neural network technique. In today’s world, handwritten document recognition is rapidly attaining the attention of researchers due to its promising behavior as assisting technology for visually impaired users. This technology is also helpful for the automatic data entry system. In the proposed system prepared a dataset of English language handwritten character images. The proposed system has been trained for the large set of sample data and tested on the sample images of user-defined handwritten documents. In this research, multiple experiments get very worthy recognition results. The proposed system will first perform image pre-processing stages to prepare data for training using a convolutional neural network. After this processing, the input document is segmented using line, word and character segmentation. The proposed system get the accuracy during the character segmentation up to 86%. Then these segmented characters are sent to a convolutional neural network for their recognition. The recognition and segmentation technique proposed in this paper is providing the most acceptable accurate results on a given dataset. The proposed work approaches to the accuracy of the result during convolutional neural network training up to 93%, and for validation that accuracy slightly decreases with 90.42%.

Offline Handwritten Character Recognition using Convolutional Neural Network

Abstract: This paper focuses on the task of recognizing handwritten Hindi characters using a Convolutional Neural Network (CNN) based. The recognized characters can then be stored digitally in the computer or used for other purposes. The dataset used is obtained from the UC Irvine Machine Learning Repository which contains 92,000 images divided into training (80%) and test set (20%). It contains different forms of handwritten Devanagari characters written by different individuals which can be used to train and test handwritten text recognizers. It contains four CNN layers followed by three fully connected layers for recognition. Grayscale handwritten character images are used as input. Filters are applied on the images to extract different features at each layer. This is done by the Convolution operation. The two other main operations involved are Pooling and Flattening. The output of the CNN layers is fed to the fully connected layers. Finally, the chance or probability score of each character is determined and the character with the highest probability score is shown as the output. A recognition accuracy of 98.94% is obtained. Similar models exist for the purpose, but the proposed model achieved a better performance and accuracy than some of the earlier models. Keywords: Devanagari characters, Convolutional Neural Networks, Image Processing

Discrete Cosine Transformation Based Approach for Offline Handwritten Character and Numeral Recognition

Handwritten Character Recognition (HCR) of Indic scripts is one of the most challenging as compared other scripts. Recognition of handwritten character is more difficult one as compares with the printed one due to their similar structure and orientation. To have a proper recognition system for handwritten characters various researchers have adapted various methods and still yet to have more feasibility and high recognition rate to achieve. Here in this proposed work we have completely focused on handwritten characters and numerals of various scripts of different regions of India such as Bengali, Odia scripts. To have a proper solution for the ambiguities that arises in handwritten one and which has been resolved using the Discrete Cosine Transformation (DCT) as feature extraction part. This transformed based featured is used to calculated GLCM values of each images of handwritten numerals and characters. To maintain the novelty of the work in simulation part we have listed up variance nature of the individual’s images. Subsequently we have also harnessed the SVM, Neural Network, and Quadratic Classifier over dataset to report the recognition rate. After validating the proposed work, we can conclude that adopted model is quite helpful in building the automatic recognition system for both handwritten character and numerals to have solution for real time problems.

Zernike Moments Based Handwritten Pashto Character Recognition Using Linear Discriminant Analysis

This paper presents an efficient Optical Character Recognition (OCR) system for offline isolated Pashto characters recognition. Developing an OCR system for handwritten character recognition is a challenging task because of the handwritten characters vary both in shape and in style and most of the time the handwritten characters also vary among the individuals. The identification of the inscribed Pashto letters becomes even palling due to the unavailability of a standard handwritten Pashto characters database. For experimental and simulation purposes a handwritten Pashto characters database is developed by collecting handwritten samples from the students of the university on A4 sized page. These collected samples are then scanned, stemmed and preprocessed to form a medium sized database that encompasses 14784 handwritten Pashto character images (336 distinguishing handwritten samples for each 44 characters in Pashto script). Furthermore, the Zernike moments are considered as a feature extractor tool for the proposed OCR system to extract features of each individual character. Linear Discriminant Analysis (LDA) is followed as a recognition tool for the proposed recognition system based on the calculated features map using Zernike moments. Applicability of the proposed system is tested by validating it with 10-fold cross-validation method and an overall accuracy of 63.71% is obtained for the handwritten Pashto isolated characters using the proposed OCR system.

Generation of Synthetic Data for Handwritten Word Alteration Detection

Fraudsters often alter handwritten contents in a document in order to achieve illicit purposes. At times, this may result in financial and mental loss to an individual or an organization. Hence, ink analysis is necessary to identify such an alteration. Convolution Neural Network (CNN) can be used to identify such cases of alteration, as CNN has emerged as a monumental success in the field of computer vision for varieties of classification tasks. But, CNN requires large amount of labeled data for training. Hence, there is a need to generate a large dataset for the experiments relating to handwritten word alteration detection. Collection, digitization, and cropping of a large number of altered and unaltered handwritten words are tedious and time consuming. To overcome such an issue, an approach for synthetic word data generation is presented in this paper for handwritten word alteration detection experiments. This scheme is designed in such a way that the synthetically generated words are very similar to the original ones. In order to achieve this, handwritten character data set is prepared using 10 blue and 10 black pens. These handwritten characters are used for creating synthetic word alteration data set. The presented approach uses relatively less number of handwritten character images to create a huge word alteration data set. Further, deep learning models are trained on the synthetically generated data set for word alteration detection.

Image iterative method for handwritten Chinese character recognition

Although accuracy and calculation time of handwritten Chinese character recognition have been greatly improved, it has not yet reached the practical application standard. Recently literatures, iterative method is used to recognize face images and good results are obtained. Therefore, image recognition is improved based on this idea. In this paper, linear combination of trigonometric functions is used as auxiliary function and image to construct discrete dynamic system. By adding bevel and shift to construct font surface and so on, the system structure is improved to solve problem of system convergence caused by large area flatness of handwritten Chinese character image, and good recognition effect is achieved. Preliminary experiments(for example, extracting 20 Chinese characters written by 9 people in the data set) are carried out, and recognition rate can reach 100% when all of them are trained, and recognition rate can reach 76.7% when each Chinese character is trained to 3 pieces; in addition, experiments were conducted on handwritten Chinese character sets changing from different angles, and the recognition rate reached 80%, which was compared with other algorithms to verify the feasibility of handwritten Chinese character recognition algorithm based on chaotic iterative trajectory characteristics of images.

Research on application of an improved deep convolutional neural network in handwritten character recognition

The classical convolutional neural network has been widely used for handwritten digit character recognition with high accuracy. However, due to its small convolutional layer, fixed size of convolution kernel and few extracted features, the recognition accuracy of complex handwritten characters are reduced. In this paper, an improved deep convolutional neural network model is proposed, which can allocate different convolution kernels according to the different information amount in the handwritten character image area for convolution, so as to better extract the effective information of the image and is more suitable for complex handwritten character recognition applications. Experiments show that the recognition rate can be higher.

Pattern Separation Network Based on the Hippocampus Activity for Handwritten Recognition

Reaching high accuracy in handwritten character recognition is an essential challenge since it is widely used in many fields such as signature analysis and forgery detection. Recently, deep learning has demonstrated efficiency in this field. The problem with deep learning is that it uses a vast number of parameters that require a large dataset for training. To overcome this problem, an intelligent network is proposed in this study, based on the computational function of the dentate gyrus of the brain’s hippocampus. The ability to separate patterns with high overlapping is a task that is referred to as the dentate gyrus. Handwritten character images have high overlapping due to various writers’ styles, or even one writer’s style under different conditions. Therefore, proposing a network based on the dentate gyrus’ functional computation can be useful in this field. One of the prominent features of the proposed network is employing two excitation steps and two inhibition steps, augmenting the accuracy of recognizing handwritten characters. The proposed network was evaluated with six datasets of digits and characters from five languages. Experiments on all of the used datasets showed promising results. Moreover, a comparative and detailed analysis of the proposed network with other SOM-based and deep learning methods is provided. Experimental results show a significant boost in accuracy. While the character error rate (CER) was smaller than 1.85% for all the experiments, the smallest CER of 0.6% was achieved by the MNIST dataset. Moreover, in recognizing patterns with high noise, the proposed network showed satisfactory results.

Handwritten Character Recognition from Images using CNN-ECOC

Recently, deep learning and character recognition have drawn the attention of many researchers. The deep neural networks have state-of-the-art performance in solving many classification and recognition problems. The Optical Character Recognition (OCR) takes an optical image of character as input and produces the corresponding character as output. It has a wide range of applications including traffic surveillance, robotics, digitization of printed articles, etc. The OCR can be implemented by using Convolutional Neural Network (CNN), which is a popular deep neural network architecture. The traditional CNN classifiers are capable of learning the important 2D features present in the images and classify them, the classification is performed by using soft-max layer. In this article, we have presented OCR by combining CNN and Error Correcting Output Code (ECOC) classifier. The CNN is used for feature extraction and the ECOC is used for classification. In order to find suitable CNN for extracting features, which can be used in combination with ECOC classifier for recognition of handwritten characters accurately, several popular CNN classifiers have been explored. The CNN-ECOC are trained and validated by using NIST handwritten character image dataset. The simulation result shows that CNN-ECOC gives higher accuracy as compared to the traditional CNN classifier.

Brahmi character recognition based on SVM (support vector machine) classifier using image gradient features

Understanding the ancient script can provide rich details of a civilization, like its cultural, political and social scenarios. Brahmi, an ancient mother script, has been a key to the development of many modern Indian scripts like Gurmukhi, Devanagari, Bangla etc. Inferring ancient writings can be a tedious job and moreover, if executed manually, it requires several language experts. The current paper presents a recognition system for Brahmi characters using linear Support Vector machine classifier. Gradient information of the character images pixels is extracted, and histogram of the gradients is stored as a feature vector for each character image. Character dataset includes both handwritten character images and images from the internet. Linear SVM classifier is trained on the feature set of 24 images of each character. The proposed recognition system is performed with an accuracy of 91.6% to recognize the Brahmi characters from the test images.

Multiobjective optimization for recognition of isolated handwritten Indic scripts

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

Off-Line character recognition using an integrated DBSCAN-ANN scheme

<p>Handwriting character recognition involves a high degree of variability and imprecision. For that, the main factor to judge the recognition accuracy is the technique that is used to extract the features. This paper developed a novel method for handwritten Arabic characters by combining the Density-Based Clustering method with statistical and morphological features. The first stage in recognition of handwritten character image has been done by binarization the image then applies noise removal techniques. The Density-Based Algorithm used to categorize and find any shape of clusters based on pixel information positions. This technique divided the image into characters. Each character will be decomposing into four regions from the centroid followed by feature extraction. These features include vertical and horizontal projections, upper and lower profile, rectangularity and orientation. The results of the present process will transfer to the Neural Network (NN) stage which generates a high level of correctness and accuracy by training. The testing results compared with two of state-of-art researches. The total accuracy of this proposed work observes a better recognition of characters.</p>

Handwritten Gurmukhi Digit Prediction using Convolutional Neural Network with Keras and Tensorflow

Intelligent Character Recognition is a term which is specifically used for the recognition of handwritten character or digit. It is a prominent research area of computer vision field of machine learning or deep learning which trained the machine to analyze the pattern of handwritten character image and identify it. Recognition of handwritten character is a hard process because single person can handwrite the same text in number of ways by making a little variation in holding the pen. Handwriting has no specific font style or size. It differs person to person or more specifically it differs how one is holding the pen. Deep Leaning has brought the breakthrough performance in this research area with its dedicated models like Convolutional Neural Network, Recurrent Neural Network etc. In this paper, we have trained model with Convolutional Neural Network with different number of layers and filters over 10,559 handwritten gurmukhi digit images and validate over 1320 images. Consequently we could achieve the maximum accuracy of 99.24%.

On developing handwritten character image database for Malayalam language script

The objective of this paper is to build a handwritten character image database for Malayalam language script. Standard handwritten document image databases are an essential requirement for the development and objective evaluation of different handwritten text recognition systems for any language script. Considerable research efforts for handwritten Malayalam character recognition are present in literature. Still, no public domain handwritten image database is available for the Malayalam language. The present work focuses on building an open source handwritten character image database for Malayalam language script. The unique orthographic representation of the Malayalam characters forms the different character classes, and the current version of the database contains 85 character classes frequently used in writing Malayalam text. Handwritten data samples collected from 77 native Malayalam writers. For extracting the character images from the handwritten data sheets, active contour model-based image segmentation algorithm utilized. Recognition experiments conducted on the created character image database by employing different feature extraction techniques. Among the considered feature descriptors, scattering convolutional network-based feature descriptors attain the highest recognition accuracy of 91.05%.