String Recognition Research Articles

Handwritten digit and Roman string recognition using gated CNN with federated learning

Handwritten digit and Roman string recognition using gated CNN with federated learning

Zombie cheminformatics: extraction and conversion of Wiswesser Line Notation (WLN) from chemical documents

PurposeWiswesser Line Notation (WLN) is a old line notation for encoding chemical compounds for storage and processing by computers. Whilst the notation itself has long since been surpassed by SMILES and InChI, distribution of WLN during its active years was extensive. In the context of modernising chemical data, we present a comprehensive WLN parser developed using the OpenBabel toolkit, capable of translating WLN strings into various formats supported by the library. Furthermore, we have devised a specialised Finite State Machine l, constructed from the rules of WLN, enabling the recognition and extraction of chemical strings out of large bodies of text. Available open-access WLN data with corresponding SMILES or InChI notation is rare, however ChEMBL, ChemSpider and PubChem all contain WLN records which were used for conversion scoring. Our investigation revealed a notable proportion of inaccuracies within the database entries, and we have taken steps to rectify these errors whenever feasible.Scientific contributionTools for both the extraction and conversion of WLN from chemical documents have been successfully developed. Both the Deterministic Finite Automaton (DFA) and parser handle the majority of WLN rules officially endorsed in the three major WLN manuals, with the parser showing a clear jump in accuracy and chemical coverage over previous submissions. The GitHub repository can be found here: https://github.com/Mblakey/wiswesser.

Enhancing handwritten numeric string recognition through incremental support vector machines

Handwritten digit recognition systems are integral to diverse applications such as postal services, banking, and document processing in our digitally-driven society. This research addresses the challenges posed by evolving datasets and dynamic scenarios in handwritten digit recognition by proposing an approach based on incremental support vector machines (ISVM). ISVM is an extension of traditional support vector machines (SVM) designed to handle scenarios where new data points become available over time. The dataset includes handwritten images (numbers “0” to “6”) and trials introducing new classes (“7”, “8”, and “9”). Evaluation utilizes k-fold cross-validation for robustness. Digital image processing involves converting images into numeric data using the histogram method. The result showed the positive outcomes of using ISVM in handwritten digit recognition, emphasizing its adaptability to incremental learning and its ability to maintain robust performance in the face of evolving datasets, which is crucial for real-world applications.

On usage of an end-to-end deep neural architecture for handwritten digit string recognition

On usage of an end-to-end deep neural architecture for handwritten digit string recognition

Residual LSTM neural network for time dependent consecutive pitch string recognition from spectrograms: a study on Turkish classical music makams

Residual LSTM neural network for time dependent consecutive pitch string recognition from spectrograms: a study on Turkish classical music makams

Handwritten Digit Recognition using Machine Learning with Python

Handwritten Digit Recognition is one of the essentially significant issues in pattern recognition applications. The main purpose of this project is to build an automatic handwritten digit recognition method for the recognition of handwritten digit strings. This paper proposes a simple convolution neural network approach to handwritten digit recognition. Convolutional Neural Network model is implemented using MNIST dataset. This dataset consists 60,000 small square 28×28pixel grayscale images of handwritten single digits between 0 and 9. The applications of digit recognition include postal mail sorting, check processing, form data entry, etc. The core of the issue exists in the capacity to foster a proficient calculation that can perceive manually written digits and which is put together by clients by the method of a scanner, tablet, and other computerized gadgets.

Tunnel Parsing with the Token’s Lexeme

Abstract The article describes a string recognition approach, engraved in the parsers generated by Tunnel Grammar Studio that use the tunnel parsing algorithm, of how a lexer and a parser can operate on the input during its recognition. Proposed is an addition of the augmented Backus-Naur form syntax that enables the formal language to be expressed with a parser grammar and optionally with an additional lexer grammar. The tokens outputted from the lexer are matched to the phrases in the parser grammar by their name and optionally by their lexeme, case sensitively or insensitively.

Automatic Botnet Attack Identification Based on Machine Learning

At present, the severe network security situation has put forward high requirements for network security defense technology. In order to automate botnet threat warning, this paper researches the types and characteristics of Botnet. Botnet has special characteristics in attributes such as packets, attack time interval, and packet size. In this paper, the attack data is annotated by means of string recognition and expert screening. The attack features are extracted from the labeled attack data, and then use K-means for cluster analysis. The clustering results show that the same attack data has its unique characteristics, and the automatic identification of network attacks is realized based on these characteristics. At the same time, based on the collection and attribute extraction of Botnet attack data, this paper uses RF, GBM, XGBOOST and other machine learning models to test the warning results, and automatically analyzes the attack by importing attack data. In the early warning analysis results, the accuracy rates of different models are obtained. Through the descriptive values of the three accuracy rates of Accuracy, Precision, and F1_Score, the early warning effect of each model can be comprehensively displayed. Among the five algorithms used in this paper, three have an accuracy rate of over 90%. The three models with the highest accuracy are used in the early warning model. The research shows that cyberattacks can be accurately predicted. When this technology is applied to the protection system, accurate early warning can be given before a network attack is launched.

Recognition of Automated Hand-written Digits on Document Images Making Use of Machine Learning Techniques

The purpose of this study is to create an automated framework that can recognize similar handwritten digit strings. For starting the experiment, the digits were separated into different numbers. The process of defining handwritten digit strings is then concluded by recognizing each digit recognition module's segmented digit. This research utilizes various machine learning techniques to produce a strong performance on the digit string recognition challenge, including SVM, ANN, and CNN architectures. These approaches use SVM, ANN, and CNN models of HOG feature vectors to train images of digit strings. Deep learning methods organize the pictures by moving a fixed-size monitor over them while categorizing each sub-image as a digit pass or fail. Following complete segmentation, complete recognition of handwritten digits is accomplished. To assess the methods' results, data must be used for machine learning training. Following that, the digit data is evaluated using the desired machine learning methodology. The Experiment findings indicate that SVM and ANN also have disadvantages in precision and efficiency in text picture recognition. Thus, the other process, CNN, performs better and is more accurate. This paper focuses on developing an effective system for automatically recognizing handwritten digits. This research would examine the adaptation of emerging machine learning and deep learning approaches to various datasets, like SVM, ANN, and CNN. The test results undeniably demonstrate that the CNN approach is significantly more effective than the ANN and SVM approaches, ranking 71% higher. The suggested architecture is composed of three major components: image pre-processing, attribute extraction, and classification. The purpose of this study is to enhance the precision of handwritten digit recognition significantly. As will be demonstrated, pre-processing and function extraction are significant elements of this study to obtain maximum consistency.

Recognition of Automated Hand-written Digits on Document Images Making Use of Machine Learning Techniques

The purpose of this study is to create an automated framework that can recognize similar handwritten digit strings. For starting the experiment, the digits were separated into different numbers. The process of defining handwritten digit strings is then concluded by recognizing each digit recognition module's segmented digit. This research utilizes various machine learning techniques to produce a strong performance on the digit string recognition challenge, including SVM, ANN, and CNN architectures. These approaches use SVM, ANN, and CNN models of HOG feature vectors to train images of digit strings. Deep learning methods organize the pictures by moving a fixed-size monitor over them while categorizing each sub-image as a digit pass or fail. Following complete segmentation, complete recognition of handwritten digits is accomplished. To assess the methods' results, data must be used for machine learning training. Following that, the digit data is evaluated using the desired machine learning methodology. The Experiment findings indicate that SVM and ANN also have disadvantages in precision and efficiency in text picture recognition. Thus, the other process, CNN, performs better and is more accurate. This paper focuses on developing an effective system for automatically recognizing handwritten digits. This research would examine the adaptation of emerging machine learning and deep learning approaches to various datasets, like SVM, ANN, and CNN. The test results undeniably demonstrate that the CNN approach is significantly more effective than the ANN and SVM approaches, ranking 71% higher. The suggested architecture is composed of three major components: image pre-processing, attribute extraction, and classification. The purpose of this study is to enhance the precision of handwritten digit recognition significantly. As will be demonstrated, pre-processing and function extraction are significant elements of this study to obtain maximum consistency.

Usage of Real Time Machine Vision in Rolling Mill

This article deals with the issue of computer vision on a rolling mill. The main goal of this article is to describe the designed and implemented algorithm for the automatic identification of the character string of billets on the rolling mill. The algorithm allows the conversion of image information from the front of the billet, which enters the rolling process, into a string of characters, which is further used to control the technological process. The purpose of this identification is to prevent the input pieces from being confused because different parameters of the rolling process are set for different pieces. In solving this task, it was necessary to design the optimal technical equipment for image capture, choose the appropriate lighting, search for text and recognize individual symbols, and insert them into the control system. The research methodology is based on the empirical-quantitative principle, the basis of which is the analysis of experimentally obtained data (photographs of billet faces) in real operating conditions leading to their interpretation (transformation into the shape of a digital chain). The first part of the article briefly describes the billet identification system from the point of view of technology and hardware resources. The next parts are devoted to the main parts of the algorithm of automatic identification—optical recognition of strings and recognition of individual characters of the chain using artificial intelligence. The method of optical character recognition using artificial neural networks is the basic algorithm of the system of automatic identification of billets and eliminates ambiguities during their further processing. Successful implementation of the automatic inspection system will increase the share of operation automation and lead to ensuring automatic inspection of steel billets according to the production plan. This issue is related to the trend of digitization of individual technological processes in metallurgy and also to the social sustainability of processes, which means the elimination of human errors in the management of the billet rolling process.

DenseNet-CTC: An end-to-end RNN-free architecture for context-free string recognition

String recognition is one of the challenging tasks in document analysis and recognition areas. Recently, with the surge of interest in end-to-end segmentation-free methods, CRNN (Convolution Recurrent Neural Network), which is a combination of CNN (Convolutional Neural Network) and RNN-CTC (Recurrent Neural Network-Connectionist Temporal Classification), has been widely applied to string recognition. However, in some context-free cases, where a character is followed by arbitrary characters like the digit string, there may be no or very few context links in these strings. In this paper, we propose a new end-to-end RNN-free architecture especially for context-free string recognition and apply it to Handwritten Digit String Recognition (HDSR) task. The proposed architecture is based on CNN and CTC, but without the usage of RNN, and we apply column-wise fully connected layers to connect the convolutional layers and CTC directly. Moreover, to compensate for the possible reduction in modeling capabilities caused by the absence of RNN, we apply densely connected convolutional layers to extract efficient features. We test this new architecture on three public HDSR benchmarks (ORAND-CAR-A, ORAND-CAR-B and CVL HDS) and three other datasets that include a handwritten telephone/postcode dataset PhPAIS and two non-Arabic digit datasets (C-Bangla and C-Hindi). Furthermore, we generate three handwritten digit string datasets to further analyze the influence of RNN. The recognition results on all datasets demonstrate the superiority of the proposed model.

A Cloud Edge Collaborative Intelligence Method of Insulator String Defect Detection for Power IIoT

Using unmanned aerial vehicles (UAVs) for equipment condition monitoring is an important application of Industrial Internet of Things (IIoT), and the limited energy is the key factor to restrict the application of UAV. In order to reduce the computational load for intelligence computing of UAV, this article proposes a cloud edge collaborative intelligent method for object detection, and applies it to insulator string recognition defect detection in the power IIoT. First, the impact of the extremely large aspect ratio of object on the detection accuracy and the computational load is analyzed, then the cloud edge collaborative intelligent method for insulator string detection and defect recognition is presented, in which on the UAV side a low cost method is proposed for estimating possible directions of insulator strings, and on the cloud side, an effective method is proposed for insulator string defect detection. The experimental results show the effectiveness of the proposed algorithm. To the best knowledge of us, this article is the first work to analyze the impact of the extremely large aspect ratio of insulator string on the detection accuracy and the computational load.

A comprehensive comparison of end-to-end approaches for handwritten digit string recognition

Over the last decades, most approaches proposed for handwritten digit string recognition (HDSR) have resorted to digit segmentation, which is dominated by heuristics, thereby imposing substantial constraints on the final performance. Few of them have been based on segmentation-free strategies where each pixel column has a potential cut location. Recently, segmentation-free strategies has added another perspective to the problem, leading to promising results. However, these strategies still show some limitations when dealing with a large number of touching digits. To bridge the resulting gap, in this paper, we hypothesize that a string of digits can be approached as a sequence of objects. We thus evaluate different end-to-end approaches to solve the HDSR problem, particularly in two verticals: those based on object-detection (e.g., Yolo and RetinaNet) and those based on sequence-to-sequence representation (CRNN).The main contribution of this work lies in its provision of a comprehensive comparison with a critical analysis of the above mentioned strategies on five benchmarks commonly used to assess HDSR, including the challenging Touching Pair dataset, NIST SD19, and two real-world datasets (CAR and CVL) proposed for the ICFHR 2014 competition on HDSR. Our results show that the Yolo model compares favorably against segmentation-free models with the advantage of having a shorter pipeline that minimizes the presence of heuristics-based models. It achieved a 97%, 96%, and 84% recognition rate on the NIST-SD19, CAR, and CVL datasets, respectively.

A comparative case study of neural network training by using frame-level cost functions for automatic speech recognition purposes in Spanish

Training procedures of a deep neural network are still an area with ample research possibilities and constant improvement either to increase its efficiency or its time performance. One of the lesser-addressed components is its objective function, which is an underlying aspect to consider when there is the necessity to achieve better error rates in the area of automatic speech recognition. The aim of this paper is to present two new variations of the frame-level cost function for training a deep neural network with the purpose of obtaining superior word error rates in speech recognition applied to a case study in Spanish. The first proposed function is a fusion between the boosted cross-entropy and the so called cross-entropy/log-posterior-ratio. The main idea is to jointly emphasize the prediction of difficult/crucial frames provided by a boosting factor and at the same time enlarge the distance between the target senone and its closest competitor. The second proposal is a fusion between the non-uniform mapped cross-entropy and the cross-entropy/log-posterior-ratio. This function utilizes both the mapped function to enhance the frames that have ambiguity in their belonging to specific senones and the log-posterior-ratio with the purpose of separating the target senone against the most competing tied tri-phone state. The proposed approaches are compared against those frame-level cost functions discussed in the state of the art. This comparative has been made by using a personalized mid-vocabulary speaker-independent voice corpus. This dataset is employed for the recognition of digit strings and personal name lists in Spanish from the northern central part of Mexico on a connected-words phone dialing task. A relative word error rate improvement of 15.14% and 12.30% is obtained with the two proposed approaches, respectively, against the plain well-established cross-entropy loss function.

Impact of literacy on the functional connectivity of vision and language related networks

Learning to read leads to functional and structural changes in cortical brain areas related to vision and language. Previous evidence suggests that the Visual Word Form Area (VWFA), a region devoted to the recognition of letter strings in literate persons, acts as an interface between both systems. While different studies have performed univariate analyses to study the effects of literacy on brain function, little is known about its impact on whole functional networks, especially when literacy is acquired during adulthood. We investigated functional connectivity in three groups of adults with different literacy status: illiterates, ex-illiterates (i.e., who learned to read during adulthood), and literates (i.e., who learned to read in childhood). We used a data-driven, multivariate whole brain approach (Independent Component Analysis [ICA]) combined with a region of interest (ROI) analysis in order to explore the functional connectivity of the VWFA with four ICA networks related to vision and language functions. ICA allowed for the identification of four networks of interest: left fronto-parietal, auditory, medial visual and lateral visual functional networks, plus a control right fronto-parietal network. We explored the effects literacy on the connectivity between the VWFA and these networks, trying furthermore to disentangle the roles of reading proficiency and age of acquisition (i.e., literacy status) in these changes. Results showed that functional connectivity between the VWFA and the left fronto-parietal and lateral visual networks increased and decreased, respectively, with literacy. Moreover, the functional coupling of the VWFA and the auditory network decreased with literacy. This study provides novel insights in the mechanisms of reading acquisition and brain plasticity, putting to light the emergence of the VWFA as a bridge between language and vision. Further studies are required to characterize the interplay of proficiency and age of reading acquisition, and its relevance to models of brain plasticity across lifespan.

Motor representations underlie the reading of unfamiliar letter combinations

Silent reading is a cognitive operation that produces verbal content with no vocal output. One relevant question is the extent to which this verbal content is processed as overt speech in the brain. To address this, we acquired sound, eye trajectories and lips’ dynamics during the reading of consonant-consonant-vowel (CCV) combinations which are infrequent in the language. We found that the duration of the first fixations on the CCVs during silent reading correlate with the duration of the transitions between consonants when the CCVs are actually uttered. With the aid of an articulatory model of the vocal system, we show that transitions measure the articulatory effort required to produce the CCVs. This means that first fixations during silent reading are lengthened when the CCVs require a greater laryngeal and/or articulatory effort to be pronounced. Our results support that a speech motor code is used for the recognition of infrequent text strings during silent reading.

Recognizing the Data Type of Firmware Data Segments With Deep Learning

Data segment analysis is of great value for firmware analysis. The data segment contains abundant information such as pointers and strings which is helpful for accelerating the process of code segment analysis. In this paper, we propose a novel approach of applying deep learning to solve the problem of data type identification in data segments, that is a fundamental problem in data segment analysis. We define 3 data types of data segment, then design several data segment byte feature extraction methods to construct feature sequences, and finally present a deep learning-based approach with feature sequences as input to recognize the data type byte by byte. Then, the recognized type can be further corrected efficiently by prior knowledge. Based on the data segment of a firmware, we built a dataset that included 18,032,352 samples (in bytes of data segment). We implement a prototype system and evaluate it with our dataset, then determine reasonable models and hyperparameters through several experiments, and eventually confirm that deep learning techniques are suitable for identifying the data type in data segment. Kappa coefficient of our data type recognition reached 0.96 and the models can be retained quickly. Using 131,072 samples in our dataset for 32 seconds of training, the accuracy can reach 90%; the accuracy can reach 97% with 273 seconds of training and 950,272 samples. Furthermore, our approach has higher accuracy than IDA in string recognition. In experiments, the recall and precision of our approach reached 96.5% and 90% respectively, whereas corresponding results of IDA is 92.9% and 85.7%. In addition, we selected 8 open source software to compile and test, and compared the detection results with TypeMiner. Experiments show that our method has certain cross-platform and operating system capabilities, and performs better than TypeMiner on some software.

基于Mask-RCNN无分割手写数字字符串的识别

基于Mask-RCNN无分割手写数字字符串的识别

HDSR-Flor: A Robust End-to-End System to Solve the Handwritten Digit String Recognition Problem in Real Complex Scenarios

Automatic handwriting recognition systems are of interest for academic research fields and for commercial applications. Recent advances in deep learning techniques have shown dramatic improvement in relation to classic computer vision problems, especially in Handwritten Text Recognition (HTR). However, several approaches try to solve the problem of deep learning applied to Handwritten Digit String Recognition (HDSR), where it has to deal with the low number of trainable data, while learning to ignore any writing symbol around the digits (noise). In this context, we present a new optical model architecture (Gated-CNN-BGRU), based on HTR workflow, applied to HDSR. The International Conference on Frontiers of Handwriting Recognition (ICFHR) 2014 competition on HDSR were used as baselines to evaluate the effectiveness of our proposal, whose metrics, datasets and recognition methods were adopted for fair comparison. Furthermore, we also use a private dataset (Brazilian Bank Check - Courtesy Amount Recognition), and 11 different approaches from the state-of-the-art in HDSR, as well as 2 optical models from the state-of-the-art in HTR. Finally, the proposed optical model demonstrated robustness even with low data volume (126 trainable data, for example), surpassing the results of existing methods with an average precision of 96.50%, which is equivalent to an average percentage of improvement of 3.74 points compared to the state-of-the-art in HDSR. In addition, the result stands out in the competition’s CVL HDS set, where the proposed optical model achieved a precision of 93.54%, while the best result so far had been from Beijing group (from the competition itself), with 85.29%.