Please use this identifier to cite or link to this item:
Title: A Hybrid Approach for Handwritten Devanagari Numerals Recognition Using HOG Algorithm and K-NN Classifier
Authors: Shukla, Bhavesh Kumar
Supervisor: Kumar, Rajiv
Keywords: OCR;HOG;HDNR;KNN;Computer Science;Computer Science Applications
Issue Date: 28-Jul-2015
Abstract: Character Recognition has become fascinating research area over a last few decades. Reading handwritten characters is easy task for human but it is difficult task for machines. Optical Character Recognition (OCR) System is used to recognize handwritten and printed characters. A lot of work has been done on English characters but these days Indian scripts have become interesting research area. Devanagari is one of them. In the present study, an effort is given for handwritten Devanagari numerals recognition. Many approaches have been developed in the field of character recognition but still it remains challenging work for researchers. In Recognition Process, Neural network is used as basic technique in most of the approaches. A lot of training and large computations are required in these approaches. So, an effort is done by author to make an easy hybrid approach for handwritten Devanagari numerals recognition (HDNR). This comes under the category of Offline recognition process. In proposed approach, various techniques such as binarization, filtering, smoothing, normalization and thinning are used in preprocessing stage. Bounding box is used for segmentation. Feature extraction is main portion of OCR system because accuracy is mostly based on extracted features. Here Histogram of Oriented Gradients (HOG) algorithm is used for feature extraction. There are a lot of feature extraction methods but we preferred HOG because of its better feature evaluation capacity. After feature extraction, they are classified into ten classes such as zero, one, two, three, four, five, six, seven, eight and nine. K-NN Classifier has been trained with these features. There are various classifiers but we selected K-NN Classifier because its accuracy is better as comparison to other classifiers. There are a lot of applications of handwritten Devanagari numerals recognition such as reading bank cheques, passport readers, postal code readers, commercial forms reader, bill processing systems etc. Proposed Hybrid approach has been applied to many documents and author obtained satisfying results.
Description: M.Tech.
Appears in Collections:Masters Theses@CSED

Files in This Item:
File Description SizeFormat 
3423.pdf2.06 MBAdobe PDFThumbnail

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.