scispace - formally typeset
Search or ask a question

Showing papers on "Intelligent word recognition published in 2002"


Journal ArticleDOI
TL;DR: This survey is divided into two parts, the first one dealing with the general aspects of Cursive Word Recognition, the second one focusing on the applications presented in the literature.

265 citations


Journal ArticleDOI
TL;DR: This analysis demonstrates how detection of various high-level features of the Bengali character set might help formulate successful multistage OCR design.

94 citations


Journal ArticleDOI
TL;DR: This work uses writer-independent writing style models (lexemes) to identify the styles present in a particular writer's training data and updates these models using the writer's data, demonstrating the feasibility of this approach on both isolated handwritten character recognition and unconstrained word recognition tasks.
Abstract: Writer-adaptation is the process of converting a writer-independent handwriting recognition system into a writer-dependent system. It can greatly increasing recognition accuracy, given adequate writer models. The limited amount of data a writer provides during training constrains the models' complexity. We show how appropriate use of writer-independent models is important for the adaptation. Our approach uses writer-independent writing style models (lexemes) to identify the styles present in a particular writer's training data. These models are then updated using the writer's data. Lexemes in the writer's data for which an inadequate number of training examples is available are replaced with the writer-independent models. We demonstrate the feasibility of this approach on both isolated handwritten character recognition and unconstrained word recognition tasks. Our results show an average reduction in error rate of 16.3 percent for lowercase characters as compared against representing each of the writer's character classes with a single model. In addition, an average error rate reduction of 9.2 percent is shown on handwritten words using only a small amount of data for adaptation.

94 citations


Proceedings ArticleDOI
06 Aug 2002
TL;DR: A new character segmentation algorithm (ACSA) of Arabic scripts is presented, which yields on the segmentation of isolated handwritten words in perfectly separated characters based on morphological rules constructed at the feature extraction phase.
Abstract: Character segmentation is a necessary preprocessing step for character recognition in many OCR systems. It is an important step because incorrectly segmented characters are unlikely to be recognized correctly. The most difficult case in character segmentation is the cursive script. The scripted nature of Arabic written language poses some high challenges for automatic character segmentation and recognition. In this paper, a new character segmentation algorithm (ACSA) of Arabic scripts is presented. The developed segmentation algorithm yields on the segmentation of isolated handwritten words in perfectly separated characters. It is based on morphological rules, which are constructed at the feature extraction phase. Finally, ACSA is combined with an existing handwritten Arabic character recognition system (RECAM).

90 citations


Journal ArticleDOI
TL;DR: A prototype of the OCR system for printed Oriya script achieves 96.3% character level accuracy on average, and the feature detection methods are simple and robust, and do not require preprocessing steps like thinning and pruning.
Abstract: This paper deals with an Optical Character Recognition (OCR) system for printedOriya script. The development of OCR for this script is difficult because a large number of character shapes in the script have to be recognized. In the proposed system, the document image is first captured using a flat-bed scanner and then passed through different preprocessing modules like skew correction, line segmentation, zone detection, word and character segmentation etc. These modules have been developed by combining some conventional techniques with some newly proposed ones. Next, individual characters are recognized using a combination of stroke and run-number based features, along with features obtained from the concept of water overflow from a reservoir. The feature detection methods are simple and robust, and do not require preprocessing steps like thinning and pruning. A prototype of the system has been tested on a variety of printed Oriya material, and currently achieves 96.3% character level accuracy on average.

81 citations


Proceedings ArticleDOI
11 Aug 2002
TL;DR: A complete system able to classify Arabic handwritten words of one hundred different writers is proposed and discussed, and successful recognition results are reported.
Abstract: Hidden Markov models (HMM) have been used with some success in recognizing printed Arabic words. In this paper, a complete scheme for totally unconstrained Arabic handwritten word recognition based on a model discriminant HMM is presented. A complete system able to classify Arabic handwritten words of one hundred different writers is proposed and discussed. The system first attempts to remove some of variation in the images that do not affect the identity of the handwritten word. Next, the system codes the skeleton and edge of the word so that feature information about the lines in the skeleton is extracted. Then a classification process based on the HMM approach is used. The output is a word in the dictionary. A detailed experiment is carried out and successful recognition results are reported.

64 citations


Proceedings ArticleDOI
06 Aug 2002
TL;DR: New methods for the creation of classifier ensembles based on feature selection algorithms are introduced, and are evaluated and compared to existing approaches in the context of handwritten word recognition, using a hidden Markov model recognizer as basic classifier.
Abstract: The study of multiple classifier systems has become an area of intensive research in pattern recognition. Also in handwriting, recognition, systems combining several classifiers have been investigated. In the paper new methods for the creation of classifier ensembles based on feature selection algorithms are introduced. These new methods are evaluated and compared to existing approaches in the context of handwritten word recognition, using a hidden Markov model recognizer as basic classifier.

59 citations


Proceedings ArticleDOI
06 Aug 2002
TL;DR: This paper introduces a framework to combine results of multiple classifiers and presents an intuitive run-time weighted opinion pool combination approach for recognizing cursive handwritten words with a large size vocabulary.
Abstract: Due to large shape variations in human handwriting, recognition accuracy of cursive handwritten word is hardly satisfying using a single classifier. In this paper we introduce a framework to combine results of multiple classifiers and present an intuitive run-time weighted opinion pool combination approach for recognizing cursive handwritten words with a large size vocabulary. The individual classifiers are evaluated run-time dynamically. The final combination is weighted according to their local performance. For an open vocabulary recognition task, we use the ROVER algorithm to combine the different strings of characters provided by each classifier. Experimental results for recognizing cursive handwritten words demonstrate that our new approach achieves better recognition performance and reduces the relative error rate significantly.

45 citations


Patent
03 Jan 2002
TL;DR: In this paper, a combined holistic and analytic recognition system was proposed to recognize an input word or phrase image by matching an input string of character features against a string of prototype features for a plurality of reference words in a lexicon.
Abstract: In a combined holistic and analytic recognition system, the holistic recognition module will recognize an input word or phrase image by matching an input string of character features for the whole word or phrase against a string of prototype features for a plurality of reference words in a lexicon. This will yield a holistic answer list of recognized word or phrase candidates for the input word or phrase along with a confidence value for each answer on the list. At the same time based on each answer in the answer list, the holistic recognition modules will generate a list of character features and segment the character features into sets of each character in an answer. The analytical recognition module uses segmentation hypotheses from the segmented character feature sets to cut the image of the input string of characters into individual character images. A plurality of character images for the various segmentation hypotheses will be recognized to produce an analytical answer list having a plurality of word or phrase answers for the input word or phrase. Each analytic word answer will have a confidence value based on the combined confidence of recognizing each character. The holistic answer list and the analytic answer list will be examined to find the best answer from the two lists as the recognition of the input handwritten text.

45 citations


01 Jan 2002
TL;DR: Novel search strategies and a novel verification approach are introduced that allow us to achieve a 120 speedup and 10% accuracy improvement over a state-of-art baseline recognition system for a very-large vocabulary recognition task (80,000 words).
Abstract: Considerable progress has been made in handwriting recognition technology over the last few years. Thus far, handwriting recognition systems have been limited to small-scale and very constrained applications where the number of different words that a system can recognize is the key point for its performance. The capability of dealing with large vocabularies, however, opens up many more applications. In order to translate the gains made by research into large and very-large vocabulary handwriting recognition, it is necessary to further improve the computational efficiency and the accuracy of the current recognition strategies and algorithms. In this thesis we focus on efficient and accurate large vocabulary handwriting recognition. The main challenge is to speedup the recognition process and to improve the recognition accuracy. However, these two aspects are in mutual conflict. It is relatively easy to improve recognition speed while trading away some accuracy. But it is much harder to improve the recognition speed while preserving the accuracy. First, several strategies have been investigated for improving the performance of a baseline recognition system in terms of recognition speed to deal with large and very-large vocabularies. Next, we improve the performance in terms of recognition accuracy while preserving all the original characteristics of the baseline recognition system: omniwriter, unconstrained handwriting, and dynamic lexicons. The main contributions of this thesis are novel search strategies and a novel verification approach that allow us to achieve a 120 speedup and 10% accuracy improvement over a state-of-art baseline recognition system for a very-large vocabulary recognition task (80,000 words). The improvements in speed are obtained by the following techniques: lexical tree search, standard and constrained lexicon-driven level building algorithms, fast two-level decoding algorithm, and a distributed recognition scheme. The recognition accuracy is improved by post-processing the list of the candidate N-best-scoring word hypotheses generated by the baseline recognition system. The list also contains the segmentation of such word hypotheses into characters. A verification module based on a neural network classifier is used to generate a score for each segmented character and in the end, the scores from the baseline recognition system and the verification module are combined to optimize performance. A rejection mechanism is introduced over the combination of the baseline recognition system with the verification module to improve significantly the word recognition rate to about 95% while rejecting 30% of the word hypotheses.

44 citations


Patent
12 Sep 2002
TL;DR: In this paper, an activity-based system and method for on-line character recognition that requires reduced amounts of memory for code or data, is alphabet-independent, and can be trained by entering the alphabet once.
Abstract: An “activity”-based system and method for on-line character recognition that requires reduced amounts of memory for code or data, is alphabet-independent, and can be trained by entering the alphabet once. The alphabet-independent nature of the algorithm, as well as the ease with which recognition may be optimized dynamically, makes it particularly well suited for writing in noisy environments (e.g., mobile or on a subway) or by persons with impaired motor skills or nervous conditions.

Proceedings ArticleDOI
11 Aug 2002
TL;DR: Two innovative techniques that contribute to the high efficiency in recognition of the mixed Chinese/English text line are presented, including a progressive search strategy based on character verification and a tree-based fast match technique with a confidence-guided adaptive stopping mechanism.
Abstract: In the past several years, we have been developing a high performance OCR engine for machine printed Chinese/English documents. We present two innovative techniques that contribute to the high efficiency in recognition of the mixed Chinese/English text line. They are (1) a progressive search strategy based on character verification, and (2) a tree-based fast match technique with a confidence-guided adaptive stopping mechanism. The efficacy of the proposed techniques is confirmed by experiments in a benchmark test.

Patent
Thomas Kwok1, Michael P. Perrone1
19 Feb 2002
TL;DR: In this paper, a handwritten word is transcribed into a list of possibly correct transcriptions of the handwritten word and a decision is made, according to a number of combination rules, as to which text word in the nearest neighbor lists or the recently transcribed list is the best transcription of handwritten word.
Abstract: A handwritten word is transcribed into a list of possibly correct transcriptions of the handwritten word. The list contains a number of text words, and this list is compared with previously stored set of lists of text words. Based on a metric, one or more nearest neighbor lists are selected from the set. A decision is made, according to a number of combination rules, as to which text word in the nearest neighbor lists or the recently transcribed list is the best transcription of the handwritten word. This best transcription is selected as the appropriate text word transcription of the handwritten word. The selected word is compared to a true transcription of the selected word. Machine learning techniques are used when the selected and true transcriptions differ. The machine learning techniques create or update rules that are used to determine which text word of the nearest neighbor lists or the recently transcribed list is the correct transcription of the handwritten word.

Proceedings ArticleDOI
11 Aug 2002
TL;DR: The primary concern of the approach is the modeling of human motor functionality while writing characters by looking at the whole pen trajectory where the time evaluation of the pen coordinates plays a crucial role.
Abstract: This paper presents the online handwriting recognition for Indian scripts. The primary concern of the approach is the modeling of human motor functionality while writing characters. This is achieved by looking at the whole pen trajectory where the time evaluation of the pen coordinates plays a crucial role. A low complexity classifier was designed and the proposed similarity measure appears to be quite robust against wide variations in writing styles. Initially, the approach was applied for online recognition of handwritten characters in Devnagari and Bangla, the two major Indian scripts. A test on a dataset of considerable size shows promising recognition rates: 97.29% for Devnagari and 96.34% for Bangla.

Journal ArticleDOI
TL;DR: An offline word-recognition system based on structural information in the unconstrained written word and a two-dimensional fuzzy word classification system where the spatial location and shape of the membership functions are derived from the training words are developed.
Abstract: This paper presents an offline word-recognition system based on structural information in the unconstrained written word. Oriented features in the word are extracted with the Gabor filters. We estimate the Gabor filter parameters from the grayscale images. A two-dimensional fuzzy word classification system is developed where the spatial location and shape of the membership functions are derived from the training words. The system achieves an average recognition rate of 74% for the word being correctly classified in the top position and an average of 96% for the word being correctly classified within the top five positions.


Proceedings ArticleDOI
06 Aug 2002
TL;DR: This paper investigates various confidence measures and their integration in an isolated word recognition system as well as in a sentence recognition system.
Abstract: In this paper we study the use of confidence measures for an on-line handwriting recognizer. We investigate various confidence measures and their integration in an isolated word recognition system as well as in a sentence recognition system. In isolated word recognition tasks, the rejection mechanism is designed in order to reject the outputs of the recognizer that are possibly wrong, which is the case for badly written words, out-of-vocabulary words or general drawing. In sentence recognition tasks, the rejection mechanism allows rejecting parts of the decoded sentence.

Proceedings ArticleDOI
07 Nov 2002
TL;DR: A very fast multi-stage algorithm for the recognition of non-Latin script that identifies not only the closet match but gives the closeness of match to all other characters in the set, which is expressed in a triangular confusion matrix.
Abstract: This paper presents a very fast multi-stage algorithm for the recognition of non-Latin script Although the examples use Arabic script, the system could be adapted in minutes to deal with any character set, in particular non-Latin characters where no commercial OCR systems are available The approach used normalises isolated characters for size and extracts an image signature based on the number of black pixels in the rows and columns of the character and compares these values to a set of signatures for typical characters of the set This technique identifies not only the closet match but gives the closeness of match to all other characters in the set, which is expressed in a triangular confusion matrix

Proceedings ArticleDOI
14 Jul 2002
TL;DR: This work examines multi-modal information retrieval from broadcast video where text can be read on the screen through OCR and speech recognition can be performed on the audio track and shows that OCR is more important that speech recognition for video retrieval.
Abstract: We examine multi-modal information retrieval from broadcast video where text can be read on the screen through OCR and speech recognition can be performed on the audio track. OCR and speech recognition are compared on the 2001 TREC Video Retrieval evaluation corpus. Results show that OCR is more important that speech recognition for video retrieval. OCR retrieval can further improve through dictionary-based post-processing. We demonstrate how to utilize imperfect multi-modal metadata results to benefit multi-modal information retrieval.

Patent
Yuji Izumi1
04 Nov 2002
TL;DR: In this paper, a handwritten character recognition apparatus performs a recognition process for a handwritten input pattern to input character codes, which is similar in shape to the handwritten input patterns, using a plurality of characters.
Abstract: A handwritten character recognition apparatus performs a recognition process for a handwritten input pattern to input character codes. The handwritten character recognition apparatus recognizes a handwritten input pattern as one pictorial symbol formed of a plurality of characters. The plurality of characters are similar in shape to the handwritten input pattern.

Journal ArticleDOI
TL;DR: A performance model is presented that views word recognition as a function of character recognition and statistically "discovers" the relation between a word recognizer and the lexicon.
Abstract: The performance of any word recognizer depends on the lexicon presented. Usually, large lexicons or lexicons containing similar entries pose difficulty for recognizers. However, the literature lacks any quantitative methodology of capturing the precise dependence between word recognizers and lexicons. This paper presents a performance model that views word recognition as a function of character recognition and statistically "discovers" the relation between a word recognizer and the lexicon. It uses model parameters that capture a recognizer's ability of distinguishing characters (of the alphabet) and its sensitivity to lexicon size. These parameters are determined by a multiple regression model which is derived from the performance model. Such a model is very useful in comparing word recognizers by predicting their performance based on the lexicon presented. We demonstrate the performance model with extensive experiments on five different word recognizers, thousands of images, and tens of lexicons. The results show that the model is a good fit not only on the training data but also in predicting the recognizers' performance on testing data.

Proceedings ArticleDOI
06 Aug 2002
TL;DR: A recognition system, based on tied-mixture hidden Markov models, for handwritten address words is described, which makes use of a language model that consists of backoff character n-grams.
Abstract: In this paper a recognition system, based on tied-mixture hidden Markov models, for handwritten address words is described, which makes use of a language model that consists of backoff character n-grams. For a dictionary-based recognition system it is essential that the structure of the address (name, street, city) is known. If the single parts of the address cannot be categorized, the used vocabulary is unknown and thus unlimited. The performance of this open vocabulary recognition using n-grams is compared to the use of dictionaries of different sizes. Especially, the confidence of recognition results and the possibility of a useful post-processing are significant advantages of language models.

Proceedings ArticleDOI
06 Aug 2002
TL;DR: This paper investigates the use of both typed and handwritten queries to retrieve handwritten documents, and a recognition-based approach reported here is novel in that it expands documents in a fashion analogous to query expansion.
Abstract: This paper investigates the use of both typed and handwritten queries to retrieve handwritten documents. The recognition-based approach reported here is novel in that it expands documents in a fashion analogous to query expansion: Individual documents are expanded using N-best lists which embody additional statistical information from a hidden Markov model (HMM) based handwriting recognizer used to transcribe each of the handwritten documents. This additional information enables the retrieval methods to be robust to machine transcription errors, retrieving documents which otherwise would be unretrievable. Cross-writer experiments on a database of 10985 words in 108 documents from 108 writers, and within-writer experiments in a probabilistic framework, on a database of 537724 words in 3342 documents from 43 writers, indicate that significant improvements in retrieval performance can be achieved. The second database is the largest database of on-line handwritten documents known to its.

Journal ArticleDOI
J. Park1
TL;DR: An adaptive handwritten word recognition method based on interaction between flexible character classification and deductive decision making is presented and the experimental result shows that the proposed method has advantages in producing valid answers using the same number of features as conventional methods.
Abstract: An adaptive handwritten word recognition method is presented. A recursive architecture based on interaction between flexible character classification and deductive decision making is developed. The recognition process starts from the initial coarse level using a minimum number of features, then increases the discrimination power by adding other features adaptively and recursively until the result is accepted by the decision maker. For the computational aspect of a feasible solution, a unified decision metric, recognition confidence; is derived from two measurements: pattern confidence, evaluation of absolute confidence using shape features, and lexical confidence, evaluation of the relative string dissimilarity in the lexicon. Practical implementation and experimental results in reading the handwritten words of the address components of US mail pieces are provided. Up to a 4 percent improvement in recognition performance is achieved compared to a nonadaptive method. The experimental result shows that the proposed method has advantages in producing valid answers using the same number of features as conventional methods.

Proceedings ArticleDOI
06 Aug 2002
TL;DR: A new database for off-line handwriting recognition, that is particularly devoted to research on bank-check recognition, up to now includes instances of isolated digits and characters, basic words of worded amounts, and signatures.
Abstract: This paper presents a new database for off-line handwriting recognition. The database, that is particularly devoted to research on bank-check recognition, up to now includes instances of isolated digits and characters, basic words of worded amounts, and signatures. Pattern images are stored using a standard image format, and hence they are easily usable by several commercial and scientific image processing packages.

Proceedings ArticleDOI
10 Dec 2002
TL;DR: In this paper a new combination method for HMM based handwritten word recognizers is introduced that combines various HMMs at a more elementary level and is experimentally demonstrated in the context of a handwritten word recognition task.
Abstract: Handwritten text recognition is one of the most difficult problems in the field of pattern recognition. The combination of multiple classifiers has been proven to be able to increase the recognition rate when compared to single classifiers. In this paper a new combination method for HMM based handwritten word recognizers is introduced. In contrast with many other multiple classifier combination schemes, where the combination takes place at the decision level, the proposed method combines various HMMs at a more elementary level. The usefulness of the new method is experimentally demonstrated in the context of a handwritten word recognition task.

Proceedings ArticleDOI
11 Aug 2002
TL;DR: Three representative modeling approaches, namely the multiple-prototype-based template matching approach, the subspace approach and the continuous density hidden Markov model approach for large vocabulary, offline recognition of handwritten Chinese characters are compared.
Abstract: We compare three representative modeling approaches, namely the multiple-prototype-based template matching approach, the subspace approach and the continuous density hidden Markov model approach for large vocabulary, offline recognition of handwritten Chinese characters. On a task of classification of 4616 handwritten Chinese characters, we evaluate and compare the strength and weakness of individual approaches in terms of the classification accuracy, the memory requirement and the computational complexity. We offer recommendations for practitioners on how to make intelligent use of these modeling approaches for different purposes in different applications.

Proceedings ArticleDOI
06 Aug 2002
TL;DR: An HMM-MLP hybrid system to recognize complex date images written on Brazilian bank cheques and introduces the concept of meta-classes of digits, which is used to reduce the lexicon size of the day and year and improve the precision of their segmentation and recognition.
Abstract: Presents an HMM-MLP hybrid system to recognize complex date images written on Brazilian bank cheques. The system first segments implicitly a date image into sub-fields through the recognition process based on an HMM-based approach. Afterwards, the three obligatory date sub-fields are processed by the system (day, month and year). A neural approach has been adopted to work with strings of digits and a Markovian strategy to recognize and verify words. We also introduce the concept of meta-classes of digits, which is used to reduce the lexicon size of the day and year and improve the precision of their segmentation and recognition. Experiments show interesting results on date recognition.

Proceedings ArticleDOI
06 Aug 2002
TL;DR: The principles of a handwritten text recognition system based on the online learning of the writer shapes are presented and the proposed scheme is shown to improve the recognition rates on a sample of fifteen writings, unknown to the system.
Abstract: Handwritten text recognition is a problem rarely studied out of specific applications for which lexical knowledge can constrain the vocabulary to a limited one. In the case of handwritten text recognition, additional information can be exploited to characterize the specificity of the writing. This knowledge can help the recognition system to find coherent solutions from both the lexical and the morphological points of view. We present the principles of a handwritten text recognition system based on the online learning of the writer shapes. The proposed scheme is shown to improve the recognition rates on a sample of fifteen writings, unknown to the system.

Proceedings ArticleDOI
06 Aug 2002
TL;DR: This paper describes how to make their recognizer compact without sacrificing too much of the recognition accuracy, and reports the results of a series of experiments that were performed to help us make a good decision when the authors face several design choices.
Abstract: We (2002) have investigate how to use Gaussian mixture continuous-density hidden Markov models (CDHMMs) for handwritten Chinese character modeling and recognition. We have identified and developed a set of techniques that can be used to construct a practical CDHMM-based off-line recognition system for a large vocabulary of handwritten Chinese characters. We have reported elsewhere the key techniques that contribute to the high recognition accuracy. In this paper we describe how to make our recognizer compact without sacrificing too much of the recognition accuracy. We also report the results of a series of experiments that were performed to help us make a good decision when we face several design choices.