Online Handwritten Recognition Research Articles

Handwriting Recognition System -- A Survey

Handwriting Recognition System has been studied in the last few decades. Many approaches are presented to recognize the hand written documents or paper. These approaches focus on how we recognize our hand written words and lining documents. Today there is no. of application area of handwriting recognition system. So an overview of hand writing recognition system and their evolution is presented by available technique with their superiorities and limitations are reviewed. So current status of handwriting recognition is focusing on off-line and online handwriting recognition system. This overview represent as an update for the state of art in the hand writing recognition field. Keywords – Handwriting recognition, offline and online handwriting recognition, segmentation, feature extraction and training.

Minimum-risk training for semi-Markov conditional random fields with application to handwritten Chinese/Japanese text recognition

Semi-Markov conditional random fields (semi-CRFs) are usually trained with maximum a posteriori (MAP) criterion which adopts the 0/1 cost for measuring the loss of misclassification. In this paper, based on our previous work on handwritten Chinese/Japanese text recognition (HCTR) using semi-CRFs, we propose an alternative parameter learning method by minimizing the risk on the training set, which has unequal misclassification costs depending on the hypothesis and the ground-truth. Based on this framework, three non-uniform cost functions are compared with the conventional 0/1 cost, and training data selection is incorporated to reduce the computational complexity. In experiments of online handwriting recognition on databases CASIA-OLHWDB and TUAT Kondate, we compared the performances of the proposed method with several widely used learning criteria, including conditional log-likelihood (CLL), softmax-margin (SMM), minimum classification error (MCE), large-margin MCE (LM-MCE) and max-margin (MM). On the test set (online handwritten texts) of ICDAR 2011 Chinese handwriting recognition competition, the proposed method outperforms the best system in competition.

Improving on-line handwritten recognition in interactive machine translation

On-line handwriting text recognition (HTR) could be used as a more natural way of interaction in many interactive applications. However, current HTR technology is far from developing error-free systems and, consequently, its use in many applications is limited. Despite this, there are many scenarios, as in the correction of the errors of fully-automatic systems using HTR in a post-editing step, in which the information from the specific task allows to constrain the search and therefore to improve the HTR accuracy. For example, in machine translation (MT), the on-line HTR system can also be used to correct translation errors. The HTR can take advantage of information from the translation problem such as the source sentence that is translated, the portion of the translated sentence that has been supervised by the human, or the translation error to be amended. Empirical experimentation suggests that this is a valuable information to improve the robustness of the on-line HTR system achieving remarkable results.

Devnagari Handwriting Recognition using STANN

approaches to recognize the handwritten characters and numerals such as online cursive handwriting and numerals recognition have been proposed. Most of them are based on neural network approaches. In this paper, a technique for continuous recognition of handwritten Devanagari characters is proposed. This approach employs a method based on Spatio-Temporal Artificial Neural Network (STANN). The proposed method is efficient in the field of online handwriting recognition because of the property of STANN to detection of spikes of the continuous input signals. The method is based on different steps of recognition. The results of signals are extracted from the input handwritten characters where the spikes signals are generated by the continuous signals of the input character. A new algorithm using STANN is proposed with results in this paper.

Effect of delayed strokes on the recognition of online Farsi handwriting

Online handwriting recognition, OHR, has gained a widespread use in everyday life. In some scripts such as Farsi and Arabic, additional strokes are written after the main stroke. These delayed strokes include dots and small signs. In this paper, the delayed strokes effect was studied from two points of views: subword modeling and lexicon reduction. The model of a subword was made of concatenating the main body model and the delayed strokes models. Hidden Markov model, HMM, was employed as a classifier. The delayed strokes of an input subword were additionally exploited to reduce the lexicon size. Our proposed method was tested on TMU-OFS dataset, including 1000 online Farsi subwords, and a recognition rate of 85.2% was achieved.

Devnagari Handwriting Recognition System using Dynamic Time Warping Algorithm

Online handwriting recognition is gaining renewed interest owing to the increase of pen computing applications and new pen input devices. The recognition of Devnagari characters is different from western handwriting recognition and poses a special challenge. This paper is an overview of the technical status and the development in online Devnagari recognition system which starts few years back. The objective of this paper is to develop a system which can recognize a handwritten Devnagari character, written with the help of stylus or a digital pen. To develop the system, dynamic time warping (DTW) technique is considered. In dynamic time warping the entered character is processed in terms of time sequences x and y as a function of time. The time sequence of the character to be recognized is compared with that of all the characters stored in a code file. The distance between them is calculated. The character from the code file which gives minimum distance is considered as the recognized character. In order to improve the recognition rate, along with DTW algorithm, additional feature vectors can be used. In this paper, line extraction algorithm is developed.

A Toolkit for Teaching Arabic Handwriting

percentage of people who produce a neat and clear handwriting is declining sharply. The traditional approach for handwriting teaching is to have a dedicated teacher for long hours of handwriting practice. Unfortunately, this is not feasible in many cases. In this paper we introduce an automated tool for teaching Arabic handwriting using tablet PCs and on-line handwriting recognition techniques. This tool can simulate the tasks performed by a human handwriting teacher of detecting the segments of hypothesized writing errors and producing instructive real time feedback to help the student to improve his handwriting quality. The tool consists of two main components, the guided writing component and the free writing component. In the guided writing mode the student is required to write over transparent images for the training examples to limit his hand movements. After the student acquires the basic skills of handwriting he can practice the free writing mode where he writes with his own style, as he usually does in his daily handwritings. The first version of the tool was tested in several schools for children with edge ranging 4-11. The results are promising and show that this tool can help students to analyze their own writing and understand how they can improve it.

Recent developments in the study of rapid human movements with the kinematic theory: Applications to handwriting and signature synthesis

Human movement modeling can be of great interest for the design of pattern recognition systems relying on the understanding of the fine motor control (such as on-line handwriting recognition or signature verification) as well as for the development of intelligent systems involving in a way or another the processing of human movements. In this paper, we briefly list the different models that have been proposed in order to characterize the handwriting process and focus on a representation involving a vectorial summation of lognormal functions: the Sigma–lognormal model. Then, from a practical perspective, we describe a new stroke extraction algorithm suitable for the reverse engineering of handwriting signals. In the following section it is shown how the resulting representation can be used to study the writer and signer variability. We then report on two joint projects dealing with the automatic generation of synthetic specimens for the creation of large databases. The first application concerns the automatic generation of totally synthetic signature specimens for the training and evaluation of verification performances of automatic signature recognition systems. The second application deals with the synthesis of handwritten gestures for speeding up the learning process in customizable on-line recognition systems to be integrated in electronic pen pads.

HMM-based online handwritten gurmukhi character recognition

This paper presents a hidden Markov model-based online handwritten character recognition for Gurmukhi script. We discuss a procedure to develop a hidden Markov model database in order to recognize Gurmukhi characters. A test with 60 handwritten samples, where each sample includes 41 Gurmukhi characters, shows a 91.95% recognition rate, and an average recognition speed of 0.112 seconds per stroke. The hidden Markov model database has been developed in XML using 5330 Gurmukhi characters. This work shall be useful to implement a hidden Markov model in online handwriting recognition and its software development.

Impact of online handwriting recognition performance on text categorization

Today, there is an increasing demand of efficient archival and retrieval methods for online handwritten data. For such tasks, text categorization is of particular interest. The textual data available in online documents can be extracted through online handwriting recognition; however, this process produces errors in the resulting text. This work reports experiments on the categorization of online handwritten documents based on their textual contents. We analyze the effect of word recognition errors on the categorization performances, by comparing the performances of a categorization system with the texts obtained through online handwriting recognition and the same texts available as ground truth. Two well-known categorization algorithms (kNN and SVM) are compared in this work. A subset of the Reuters-21578 corpus consisting of more than 2,000 handwritten documents has been collected for this study. Results show that classification rate loss is not significant, and precision loss is only significant for recall values of 60–80% depending on the noise levels.

The study of mathematical expression recognition and the embedded system design

Mathematical expression recognition is an important problem about pattern recognition, because mathematical expression is an essential part of scientific literature. This paper focused on two aspects, first of all, About the recognition of on-line handwritten mathematical expressions, this paper elaborates the structural analysis strategy from the perspective of the general mathematical expression, we bring forward a feasible solution to the overall structural analysis by using some pretreatment methods, such as dominating relationships, hard-regulations, function merging, and so on. The results show that this method can overcome some shortcomings of traditional structure analysis methods, and also it can dispose some recognition errors that often appear in the space structure and logical structure of the on-line handwriting recognition system effectively, such as symbol overlapping and handshaking, complex layer structure, the table structure and so on. Secondly, on this basis, We have contrived an online handwriting recognizing system of mathematical expression based on an embedded equipment, which supports in the use of the languages of the minority, such as Uyghur and Kazakh languages .For the features of mathematical expression recognizing, and from the point of ensuring the system’s efficiency , this paper has constructed an embedded platform, on which the capability of the system has been tested, and it has compared the data gained from the test with those in PC , which offers a gist to improve the algorithm and perfect the system.

LAMPS: A sketch recognition-based teaching tool for Mandarin Phonetic Symbols I

The non-Romanized Mandarin Phonetic Symbols I (MPS1) system is a highly advantageous phonetic system for native English users studying Chinese Mandarin to learn, yet its steep initial learning curve discourages language programs to instead adopt Romanized phonetic systems. Computer-assisted language instruction (CALI) can greatly reduce this learning curve, in order to enable students to sooner benefit from the long-term advantages presented in MPS1 usage during the course of Chinese Mandarin study. Unfortunately, the technologies surrounding existing online handwriting recognition algorithms and CALI applications are insufficient in providing a “dynamic” counterpart to traditional paper-based workbooks employed in the classroom setting. In this paper, we describe our sketch recognition-based LAMPS system for teaching MPS1 by emulating the naturalness and realism of paper-based workbooks, while extending their functionality with human instructor-level critique and assessment at an automated level.

FEATURE SELECTION FOR HMM AND BLSTM BASED HANDWRITING RECOGNITION OF WHITEBOARD NOTES

In this paper, we describe feature selection experiments for online handwriting recognition. We investigated a set of 25 online and pseudo-offline features to find out which features are important and which features may be redundant. To analyze the saliency of the features, we applied a sequential forward and a sequential backward search on the feature set. A hidden Markov model and a neural network based recognizer have been used as recognition engines. In our experiments, we obtained interesting results. Using a set of only five features, we achieved a performance similar to that of the reference system that uses all 25 features. The five selected features have a low correlation and have been the top choices during the first iterations of the forward search with both recognizers. Furthermore, for both recognizers, subsets have been identified that outperform the reference system with statistical significance. In order to assess the results more rigorously, we have compared our recognizer with the widely used commercial recognizer from Microsoft.

On-line Arabic handwriting recognition with templates

After a long period of focus on western and East Asian scripts there is now a general trend in the on-line handwriting recognition community to explore recognition of other scripts such as Arabic and various Indic scripts. One difficulty with the Arabic script is the number and position of diacritic marks associated to Arabic characters. This paper explores the application of a template matching scheme to the recognition of Arabic script with a novel algorithm for dynamically treating the diacritical marks. Template based systems are robust to conditions with scarce training data and in experiments the proposed system outperformed a reference system based on the promising state-of-the-art network technique of BLSTM. Experiments have been conducted in an environment similar to that of many handheld devices with promising results both in terms of memory consumption and response time.

Combining diverse on-line and off-line systems for handwritten text line recognition

In this paper we present a multiple classifier system (MCS) for on-line handwriting recognition. The MCS combines several individual recognition systems based on hidden Markov models (HMMs) and bidirectional long short-term memory networks (BLSTM). Beside using two different recognition architectures (HMM and BLSTM), we use various feature sets based on on-line and off-line features to obtain diverse recognizers. Furthermore, we generate a number of different neural network recognizers by changing the initialization parameters. To combine the word sequences output by the recognizers, we incrementally align these sequences using the recognizer output voting error reduction framework (ROVER). For deriving the final decision, different voting strategies are applied. The best combination ensemble has a recognition rate of 84.13%, which is significantly higher than the 83.64% achieved if only one recognition architecture (HMM or BLSTM) is used for the combination, and even remarkably higher than the 81.26% achieved by the best individual classifier. To demonstrate the high performance of the classification system, the results are compared with two widely used commercial recognizers from Microsoft and Vision Objects.

A large margin approach for writer independent online handwriting classification

This paper proposes a new approach for classifying multivariate time-series with applications to the problem of writer independent online handwritten character recognition. Each time-series is approximated by a sum of piecewise polynomials in a suitably defined Reproducing Kernel Hilbert Space (RKHS). Using the associated kernel function a large margin classification formulation is proposed which can discriminate between two such functions belonging to the RKHS. The associated problem turns out to be an instance of convex quadratic programming. The resultant classification scheme applies to many time-series discrimination tasks and shows encouraging results when applied to online handwriting recognition tasks.

MIXED DISCRIMINANT TRAINING OF HYBRID ANN/HMM SYSTEMS FOR ONLINE HANDWRITTEN WORD RECOGNITION

Online handwritten word recognition systems usually rely on Hidden Markov Models (HMMs), which are effective under many circumstances, but suffer some major limitations in real world applications. Artificial neural networks (ANN) appear to be a promising alternative, however they failed to model sequence data such as online handwriting due to their variable lengths. As a consequence, by combining HMMs and ANN, we can expect to take advantage of the robustness and flexibility of the HMMs generative models and of the discriminative power of the ANN. Training such a hybrid system is not straightforward, this is why so few attempts are encountered in literature. We compare several different training schemes: maximum likelihood (ML) and maximum mutual information (MMI) criteria in the framework of online handwriting recognition with a global optimization approach defined at the word level. A new generic criterion mixing generative model and discriminant trainings is proposed, it allows to train a multistate TDNN-HMM system directly at the word level. This architecture is based on an analytical approach with an implicit segmentation. To control the implicit segmentation and to initialize correctly the system without bootstrapping with another recognition system, we have defined a process that constraints the segmentation path and a measure called Average Segmentation Rate (ASR). Recognition experiments on the online IRONOFF database demonstrated the interest of the generic training criterion and the control of the implicit segmentation.

Minimum classification error training for online handwriting recognition

This paper describes an application of the Minimum Classification Error (MCE) criterion to the problem of recognizing online unconstrained-style characters and words. We describe an HMM-based, character and word-level MCE training aimed at minimizing the character or word error rate while enabling flexibility in writing style through the use of multiple allographs per character. Experiments on a writer-independent character recognition task covering alpha-numerical characters and keyboard symbols show that the MCE criterion achieves more than 30 percent character error rate reduction compared to the baseline Maximum Likelihood-based system. Word recognition results, on vocabularies of 5k to 10k, show that MCE training achieves around 17 percent word error rate reduction when compared to the baseline Maximum Likelihood system.

Writer Independent Online Handwritten Character Recognition Using a Simple Approach

This study describes the simple approach involved in online handwriting recognition. Conventionally, the data obtained needs a lot of preprocessing including filtering, smoothing, slant removing and size normalization before recognition process. Instead of doing such lengthy preprocessing, this study presents a simple approach to extract the useful character information. The whole process requires no preprocessing and size normalization. This research evaluates the use of the Back-propagation Neural Network (BPN) and presents feature extraction mechanism in full detail to work with on-line handwriting recognition. The obtained recognition rates were 51 to 83% using the BPN for different sets of character samples. This study also describes a performance study in which a recognition mechanism with multiple thresholds is evaluated for back-propagation architecture. The results indicate that the application of multiple thresholds has significant effect on recognition mechanism. This is a writer-independent system and the method is applicable for off-line character recognition as well. The technique is tested for upper-case English alphabets for a number of different styles from different subjects.

Clustering-based Approach with Dynamic Time Warping for Thai On-line Handwriting Recognition

Clustering-based Approach with Dynamic Time Warping for Thai On-line Handwriting Recognition