Intelligent word recognition

About: Intelligent word recognition is a research topic. Over the lifetime, 2480 publications have been published within this topic receiving 45813 citations.

Offline handwritten text recognition using support vector machines

Abstract: Comenia script is a novel handwritten text introduced at primary schools in the Czech Republic This paper describes a method for handwritten text recognition (HWR) of this font In particular it proposes a method for preprocessing and normalization of data and optical character recognition based on SVM classifier We have trained and statistically evaluated several models, where we have focused on recognition of different styles of writing of the same characters — for the forensic purposes and identification of the author of a document The best model has achieved 9286 % accuracy without any further postprocessing, eg a spellchecker We also proposed using more than one classification model for character recognition that has shown to increase accuracy when compared to a single model approach

Unconstrained Japanese address recognition using a combination of spatial information and word knowledge

Abstract: We describe a new handwritten address recognition method which can correct the errors occurring in line extraction, character segmentation, and character recognition as a possible means of avoiding the error accumulation which occurs during the recognition sequence in conventional methods. We formulate the address recognition method as a minimum cost search problem. We define the character recognition cost which estimates the reliability of the character recognition result, the arrangement cost which estimates the plausibility of the character string's spatial arrangement, and the word knowledge cost which estimates the plausibility of the linguistic conditions. By using a combination of these costs, the proposed method can recognize an address which has not been extracted as a single line from input images by a conventional method. The efficiency of the proposed method is evaluated through an experiment using 600 Japanese mail images. An address recognition rate of 79.38% was obtained.

An Efficient Fuzzy Method for Bangla Handwritten Numerals Recognition

Abstract: The object of the handwritten character recognition is the recognition of data that describe handwritten objects On-line handwritten recognition deals with a time ordered sequence of data This paper presents an on-line handwritten recognition system for recognizing the Bangla numerals using the fuzzy rule-base Fuzzy logic has proved to be a powerful tool to represent imprecise and irregular patterns The selection of the representative features, which describe the shapes and location of segments, is the core of the proposed approach This paper describes a method that can be used to generate a fuzzy value database that describes the characters written by different individuals

A class-modular GLVQ ensemble with outlier learning for handwritten digit recognition

Abstract: A class-modular generalized learning vector quantization (GLVQ) ensemble method with outlier learning for handwritten digit recognition is proposed. A GLVQ classifier is a discriminative method. Though discriminative classifiers have the remarkable ability to solve character recognition problems, they are poor at outlier resistance. To overcome this problem, a GLVQ classifier trained with both digit images and outlier images is introduced. Moreover, the original 10-classification problem is separated into ten 2-classification problems using ten GLVQ classifiers, each of which recognizes its corresponding digit class. Experimental results of handwritten digit recognition and outlier rejection reveal that our method is far more superior at outlier resistance than a conventional GLVQ classifier, while maintaining its digit recognition performance.

The same is not the same - postcorrection of alphabet confusion errors in mixed-alphabet OCR recognition

Abstract: Character sets for Eastern European languages typically contain symbols that are optically almost or fully identical to Latin letters. When scanning documents with mixed Cyrillic-Latin or Greek-Latin alphabets, even high-quality OCR-software is often not able to correctly separate between Cyrillic (Greek) and Latin symbols. This effect leads to an error rate that is far beyond the usual error rates observed when recognizing single-alphabet documents. In this paper we first survey similarities between Latin and Cyrillic (Greek) letters and words for distinct languages and fonts. After briefly introducing a new and public corpus collected by our groups for evaluating OCR-technology over mixed-alphabet documents, we describe how to adapt general algorithms and tools for postcorrection of OCR results to the new context of mixed-alphabet recognition. Experimental results on Bulgarian documents from the corpus and from other sources demonstrate that a drastic reduction of error rates can be achieved.