In this paper, word recognition using neural network is proposed. Recognition process is started with the partitioning of document image into lines, words, and characters and then capturing the local features of segmented characters. After classifying the characters, the word image is transferred into unique code based on character code. This code ideally describes any form of word including word with mixed styles and different sizes. Sequence of character codes of the word form input pattern and word code is a target value of the pattern. Neural network is used to train the patterns of the words. Trained network is tested with word patterns and is recognized or unrecognized based on the network error value. Experiments have been conducted with a local database to evaluate the performance of the word recognizing system and obtained good accuracy. This method can be applied for any language word recognition system as the training is based on only unique code of the characters and words belonging to the language.
Feature Extraction plays most crucial and important role in character recognition. The selection of stable and representative set of features is the main problem in pattern recognition. Because of font characteristics and style variation of machine printed Tamil characters, feature extraction remains a problem. Feature extraction involves reducing the amount of resources required to describe a set of data. In this paper, new method has been proposed to extract structural features from Machine printed Tamil characters using horizontal and vertical projections. Based on the structural properties of upper and lower modifiers, characters are divided into various categories and features are extracted accordingly. The extracted features from the real life degraded documents are classified to identify the characters. The system has been tested with printed Tamil characters and achieves 99.67% character recognition accuracy on average. Experimental results show that structure and category of the characters are identified by the proposed method for the regular characters of various sizes.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
customersupport@researchsolutions.com
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.