Showing papers on "Intelligent word recognition published in 1995"

Comparison of learning algorithms for handwritten digit recognition

Abstract: COMPARISON OF LEARNINGALGORITHMS FOR HANDWRITTEN DIGITRECOGNITIONY. LeCun, L. Jackel, L. Bottou, A. Brunot, C. Cortes,J. Denker, H. Drucker, I. Guyon, U. M uller,E. Sackinger, P. Simard, and V. VapnikBell Lab oratories, Holmdel, NJ 07733, USAEmail: yann@research.att.comAbstractThis pap er compares the p erformance of several classi er algorithmson a standard database of handwritten digits. We consider not only rawaccuracy, but also rejection, training time, recognition time, and memoryrequirements.1

Learning algorithms for classification: A comparison on handwritten digit recognition

Abstract: This paper compares the performance of several classi er algorithms on a standard database of handwritten digits. We consider not only raw accuracy, but also training time, recognition time, and memory requirements. When available, we report measurements of the fraction of patterns that must be rejected so that the remaining patterns have misclassi cation rates less than a given threshold.

Competition and segmentation in spoken word recognition

Abstract: Spoken utterances contain few reliable cues to word boundaries, but listeners nonetheless experience little difficulty identifying words in continuous speech. The authors present data and simulations that suggest that this ability is best accounted for by a model of spoken-word recognition combining competition between alternative lexical candidates and sensitivity to prosodic structure. In a word-spotting experiment, stress pattern effects emerged most clearly when there were many competing lexical candidates for part of the input. Thus, competition between simultaneously active word candidates can modulate the size of prosodic effects, which suggests that spoken-word recognition must be sensitive both to prosodic structure and to the effects of competition. A version of the Shortlist model (D. G. Norris, 1994b) incorporating the Metrical Segmentation Strategy (A. Cutler & D. Norris, 1988) accurately simulates the results using a lexicon of more than 25,000 words.

Models of continuous speech recognition and the contents of the vocabulary

Abstract: Several models of spoken word recognition postulate that recognition is achieved via a process of competition between lexical hypotheses. Competition not only provides a mechanism for isolated word recognition, it also assists in continuous speech recognition, since it offers a means of segmenting continuous input into individual words. We present statistics on the pattern of occurrence of words embedded in the polysyllabic words of the English vocabulary, showing that an overwhelming majority (84%) of polysyllables have shorter words embedded within them. Positional analyses show that these embeddings are most common at the onsets of the longer word. Although both phonological and syntactic constraints could rule out some embedded words, they do not remove the problem. Lexical competition provides a means of dealing with lexical embedding. It is also supported by a growing body of experimental evidence. We present results which indicate that competition operates both between word candidates that...

Real-time on-line unconstrained handwriting recognition using statistical methods

Abstract: We address the problem of automatic recognition of unconstrained handwritten text. Statistical methods, such as hidden Markov models (HMMs) have been used successfully for speech recognition and they have been applied to the problem of handwriting recognition as well. We discuss a general recognition system for large vocabulary, writer independent, unconstrained handwritten text. "Unconstrained" implies that the user may write in any style e.g. printed, cursive or in any combination of styles. This is more representative of typical handwritten text where one seldom encounters purely printed or purely cursive forms. Furthermore, a key characteristic of the system is that it performs recognition in real-time on 486 class PC platforms without the large amounts of memory required for traditional HMM based systems. We focus mainly on the writer independent task. Some initial writer dependent results are also reported. An error rate of 18.9% is achieved for a writer-independent 21,000 word vocabulary task in the absence of any language models.

Character recognition without segmentation

Abstract: A segmentation-free approach to OCR is presented as part of a knowledge-based word interpretation model. It is based on the recognition of subgraphs homeomorphic to previously defined prototypes of characters. Gaps are identified as potential parts of characters by implementing a variant of the notion of relative neighborhood used in computational perception. Each subgraph of strokes that matches a previously defined character prototype is recognized anywhere in the word even if it corresponds to a broken character or to a character touching another one. The characters are detected in the order defined by the matching quality. Each subgraph that is recognized is introduced as a node in a directed net that compiles different alternatives of interpretation of the features in the feature graph. A path in the net represents a consistent succession of characters. A final search for the optimal path under certain criteria gives the best interpretation of the word features. Broken characters are recognized by looking for gaps between features that may be interpreted as part of a character. Touching characters are recognized because the matching allows nonmatched adjacent strokes. The recognition results for over 24,000 printed numeral characters belonging to a USPS database and on some hand-printed words confirmed the method's high robustness level. >

Segmentation-free word recognition with application to Arabic

Abstract: This paper describes the design and implementation of a system that recognizes machine-printed Arabic words without prior segmentation. The technique is based on describing symbols in terms of shape primitives. At recognition time, the primitives are detected on a word image using mathematical morphology operations. The system then matches the detected primitives with symbol models. This leads to a spatial arrangement of matched symbol models. The system conducts a search in the space of spatial arrangements of models and outputs the arrangement with the highest posterior probability as the recognition of the word. The advantage of using this whole word approach versus a segmentation approach is that the result of recognition is optimized with regard to the whole word. Results of preliminary experiments using a lexicon of 42,000 words show a recognition rate of 99.4% for noise-free text and 73% for scanned text.

Visual Word Recognition: An Overview

Abstract: Visual word recognition is the aspect of reading for which the use of the visual perceptual channel has the largest impact on the comprehension process. This chapter provides an overview of current research in this area, focusing on issues concerning the types of processing mechanisms and knowledge representations that are involved. Whereas recognizing and pronouncing letter patterns are specific to this medium, most of the other capacities that are utilized in understanding a text are also used in comprehending spoken language. Thus, it is within the domain of word recognition that most of what is specific to reading comprehension rather than a reflection of more general linguistic factors, including the bases of reading-specific deficits and individual differences, is found. The chapter discusses that visual word recognition exists only because of the invention of writing systems, a relatively recently development in the span of human history. Reading exploits perceptual and cognitive capacities that did not evolve specifically for this function and provides an interesting domain in which to explore them.

The handwritten trie: indexing electronic ink

Abstract: The emergence of the pen as the main interface device for personal digital assistants and pen-computers has made handwritten text, and more generally ink, a first-class object. As for any other type of data, the need of retrieval is a prevailing one. Retrieval of handwritten text is more difficult than that of conventional data since it is necessary to identify a handwritten word given slightly different variations in its shape. The current way of addressing this is by using handwriting recognition, which is prone to errors and limits the expressiveness of ink. Alternatively, one can retrieve from the database handwritten words that are similar to a query handwritten word using techniques borrowed from pattern and speech recognition. In particular, Hidden Markov Models (HMM) can be used as representatives of the handwritten words in the database. However, using HMM techniques to match the input against every item in the database (sequential searching) is unacceptably slow and does not scale up for large ink databases. In this paper, an indexing technique based on HMMs is proposed. The new index is a variation of the trie data structure that uses HMMs and a new search algorithm to provide approximate matching. Each node in the tree contains handwritten letters, where each letter is represented by an HMM. Branching in the trie is based on the ranking of matches given by the HMMs. The new search algorithm is parametrized so that it provides means for controlling the matching quality of the search process via a time-based budget. The index dramatically improves the search time in a database of handwritten words. Due to the variety of platforms for which this work is aimed, ranging from personal digital assistants to desktop computers, we implemented both main-memory and disk-based systems. The implementations are reported in this paper, along with performance results that show the practicality of the technique under a variety of conditions.

Handwritten word recognition for real-time applications

Abstract: A fast handwritten word recognition system for real time applications is presented. Preprocessing, segmentation and feature extraction are implemented using chain code representation. Dynamic matching between each character of a lexicon entry and segment(s) of input word image is used for ranking words in the lexicon. Speed of the entire recognition process is about 200 msec on a single SPARC-10 platform for lexicon size of 10. A top choice performance of 96% is achieved on a database of postal words captured at 212 dpi.

Graph-based handwritten digit string recognition

Abstract: The article presents a handwritten digit string recognition algorithm based on matching input subgraphs with prototype symbol graphs. The article defines a set of acceptable graph transformations corresponding to typical variations of the handwritten symbols. The search for a match between the input subgraph and prototype graph is conducted using this set of transformations. This approach allows us to solve the problems of structure recognition methods caused by a high variability of handwritten symbol topology. The article presents experimental results of the handwritten digit string recognition system.

Keyword spotting via word shape recognition

Abstract: With the advent of on-line access to very large collections of document images, electronic classification into areas of interest has become possible. A first approach to classification might be the use of OCR on each document followed by analysis of the resulting ASCII text. But if the quality of a document is poor, the format unconstrained, or time is critical, complete OCR of each image is not appropriate. An alternative approach is the use of word shape recognition (as opposed to individual character recognition) and the subsequent classification of documents by the presence or absence of selected keywords. Use of word shape recognition not only provides a more robust collection of features but also eliminates the need for character segmentation (a leading cause of error in OCR). In this paper we describe a system we have developed for the detection of isolated words, word portions, as well as multi-word phrases in images of documents. It is designed to be used with large, changeable, keyword sets and very large document sets. The system provides for automated training of desired keywords and creation of indexing filters to speed matching.© (1995) COPYRIGHT SPIE--The International Society for Optical Engineering. Downloading of the abstract is permitted for personal use only.

Method of registering a character pattern into a user dictionary and a character recognition apparatus having the user dictionary

Abstract: In the present invention, an input handwritten character pattern is subjected to character recognition processing, and a recognition reliability of the character as a standard characteristic feature pattern is determined from the recognition result. If the recognition reliability is low, a warning is issued. In response to the warning, a user or operator can decide whether the character pattern should be registered in the user dictionary (106). If it is decided that the character pattern should be registered in the user dictionary, the character pattern is stored in the user dictionary with the information representing that the character pattern has low recognition reliability. When character patterns registered in the user dictionary are displayed on a screen, these characters are displayed in such a manner that it is possible to distinguish characters having low recognition reliability from characters having high recognition reliability. There is also provided a user name index file (5309) for storing information regarding characteristic features of a handwritten character pattern peculiar to a specific user. Furthermore, there is also provided a password input-and-decision part (5103) for making a decision of whether or not allow to access to the user dictionary based on the information of the handwritten character pattern input by a specific user.

The analysis of error in an on-line recognition system of Arabic handwritten characters

Abstract: In this paper we describe a system that recognizes on-line Arabic handwriting characters. In this system, a dynamic programming algorithm is implemented. We present here the effects of some parameters of the system on its performances and on its recognition errors. This study shows a way to select the optimum values for some key parameters of the system to obtain minimum recognition error rates.

Handprinted word recognition on a NIST data set

Abstract: An approach to handprinted word recognition is described. The approach is based on the use of generating multiple possible segmentations of a word image into characters and matching these segmentations to a lexicon of candidate strings. The segmentation process uses a combination of connected component analysis and distance transform-based, connected character splitting. Neural networks are used to assign character confidence values to potential character within word images. Experimental results are provided for both character and word recognition modules on data extracted from the NIST handprinted character database.

Fuzzy Full-Text Searches in OCR Databases

Abstract: Though the quality of optical character recognition software is steadily improving, it is still far from being perfect. As a result, full-text databases that are lled by means of OCR software contain many errors. These errors have to be taken into consideration if such kind of databases are examined by means of full-text searches. In this chapter, we will illustrate some of the possible methods that { to a certain extent { cope with the uncertainty of the database entries. These methods add fuzziness to precisely formulated queries in order to increase their recall. In addition, the described methods are compared to the method of matching query terms exactly: the preliminary results of tests that show their eeects on recall and precision are given.

Visual inter-word relations and their use in OCR postprocessing

Abstract: A technique is presented that uses visual relationships between word images in a document to improve the recognition of the text it contains. This technique takes advantage of the visual relationships between word images that are usually lost in most conventional optical character recognition (OCR) techniques. The visual relations are defined to be the equivalence that exists between images of the same word or portions of word images. An algorithm is presented that calculates these relationships in a document. The resulting clusters are integrated with the recognition results provided by an OCR system. Inconsistencies in OCR results between equivalent images are identified and used to improve recognition performance. Experimental results are presented in which the input is provided directly from a commercial OCR system.

Segmentation and recognition of handwritten characters using subspace method

Abstract: Segmentation of characters freely written on papers is a difficult problem for a computer system. Conventionally this problem has been dealt with by image processing such os horizontal or vertical projection. But it sometimes splits and merges the character images and fails to correctly segment them, due to its lack of character recognition ability in the segmentation process. We propose in this paper a method to solve this problem by performing character recognition in segmentation process based on a subspace method. At first, a binary image on which characters are written, is scanned by a fixed sale of a window. At every scanning location, 196(7/spl times/7/spl times/4) features are obtained and projected to each character subspace. The character recognition using subspace method is carried out and character name (or group name) and its confidence are obtained. Since this character segmentation based on the subspace method performs the character recognition simultaneously, it can be applied to isolatedly or cursively written characters.

Recognition of chain-coded handwritten character images with scanning n-tuple method

Abstract: A method of applying n-tuple recognition techniques to handwritten OCR, which involves scanning an n-tuple classifier over a chain-code of the image, is described. The traditional advantages of n-tuple recognition, i.e. training and recognition speed, are retained, while offering superior recognition accuracy, as demonstrated by results on three widely used data sets.

Computer recognition of printed Bangla script

Abstract: This paper considers optical character recognition (OCR) of Bangla, the second most popular script in the Indian subcontinent. A complete OCR system is described for documents of single Bangla font, where more than three hundred character shapes are recognized by a combination of template and feature-matching approach. Here the document image captured by a flatbed scanner is subject to tilt correction, line, word and character segmentation, simple and compound character separation, feature extraction and finally character recognition. Some character occurrence statistics have been computed to aid the recognition process. The simple character recognition is done by a feature-based tree classifier, and the compound character recognition involves a template matching approach preceded by a feature-based grouping. At present, recognition accuracy of about 96% is obtained by the system.

On-Line Handwritten Alphanumeric Character Recognition Using Feature Sequences

Abstract: In this paper we present an approach in which an on-line handwritten character is characterized by a sequence of dominant points in strokes and a sequence of writing directions between consecutive dominant points. The directional information is used for character preclassification and the positional information is used for fine classification. Doth preclassification and fine classification are based on dynamic programming matching. A recognition experiment has been conducted with 62 character classes of different writing styles and 21 people as data contributors. The recognition rate of this experiment is 91%, with 7.9% substitution rate and 1.1% rejection rate. The average processing time is 0.35 second per character on a 486 50MHz personal computer.

An HMM-based legal amount field OCR system for checks

Abstract: The system described in this paper applies hidden Markov technology to the task of recognizing the handwritten legal amount on personal checks. We argue that the most significant source of error in handwriting recognition is the segmentation process. In traditional handwriting OCR systems, recognition is performed at the character level, using the output of an independent segmentation step. Using a fixed stepsize series of vertical slices from the image, the HMM system described in this paper avoids taking segmentation decisions early in the recognition process.

Lexicon-driven word recognition

Abstract: Most conventional document understanding systems use lexicons only in a postprocessing step to verify or correct character recognition results. The authors present a new approach to word recognition that uses a lexicon to "drive" the recognition process. Lexicon words are encoded in trie data structures, and recognition of a word image is done by searching a lexicon trie for a path whose node characters yield the best match to the word image. This approach has two important advantages. First, it is segmentation-free; there is no need to presegment the text image into isolated characters. Second, it performs recognition by verifying character hypotheses, as opposed to the classification method used in most conventional optical character recognition (OCR) systems. Hence, the recognition process is more efficient and the results are more accurate. They demonstrated the feasibility and the advantage of this approach with a lexicon size of more than 50000 words, on severely degraded images.

System for indexing document images

Abstract: When texts recognized by an OCR are registered and those texts are searched by a search word, a state in which the search cannot be performed depending on an error recognition at the time of the recognition by the OCR is eliminated. It is an object of the invention to realize a process such that no burden is exerted on an operator or an apparatus by the above state. There are provided an OCR processor for recognizing stored image information and outputting a recognition result while switching the number of candidate characters to be outputted as a recognition result in accordance with a degree of a likelihood; and a document searcher for forming character trains for search from the recognition result and for registering as a search file.

Device and method for handwriting recognition with adaptive weighting of recognition data

Abstract: An adaptive weighting handwriting recognition device and method compares information representing handwritten input with stored recognition data at least some of which has a weighting value associated therewith. The weighting values remains fixed during comparison of the handwritten input and stored recognition data and provide candidate recognition information, which can be further processed with user editing instructions to modify and correct the candidate recognition information. During user editing, the weighting values of the stored recognition data associated with the corrected candidate recognition information modified are varied to enhance the likelihood of correct future handwriting recognition.

Realization of a high-performance bilingual Chinese-English OCR system

Abstract: This paper focuses on the realization of a bilingual Chinese-English OCR system. First, the Twice-Segment Algorithm is used for segmentation of documents with Chinese and English characters mixed. Then the comprehensive recognition method is employed to improve the robustness of Chinese character recognition. A new measurement of robustness of OCR recognition performance is also put forward here. Finally, exciting experimental results are given.