Please use this identifier to cite or link to this item: https://dl.ucsc.cmb.ac.lk/jspui/handle/123456789/3179
Title: Off-Line Handwriting Recognition System for Identifying City Names in Sri Lanka
Authors: Ekanayaka, E.M.A.K.
Issue Date: 30-Jun-2015
Abstract: Handwriting recognition remains offline community zone open research. And many researchers take attention to this area because it is linked with many applications. This thesis considers offline handwritten Sinhala name of the postal city recognition using free segmentation algorithms. Four phases are performed in this research as pre-processing, feature detection, recognition and post-processing. There are seven steps in the pre-processing phase such as noise removing, thresholding, detection of the rectangular area, skew detecting, skew correction, underline removal and thinning. Feature detection techniques used here are the horizontal histogram profile projection, the profile of the vertical projection histogram and Gabor filter. Thirteen features are identified for each image using these techniques. A neural network is used the feed forward with a hidden layer. This is based on the supervised learning algorithm. There are six output nodes to classify fifty different postal towns. The data for the test set and training set contains city names after writing Sinhala people with real purpose, as well as students and others. The well-trained neural network outputs a binary number corresponding to the name of the recognized city. The post-processing technique uses a Microsoft Access database as the decimal value of the binary number in the name of the city is allocated. It gives the recognized city name in a user readable manner. The name of the city recognition accuracy of the system is about 41%. And the system is implemented using Matlab 7.0.
URI: http://hdl.handle.net/123456789/3179
Appears in Collections:Master of Computer Science - 2015

Files in This Item:
File Description SizeFormat 
12440213.pdf
  Restricted Access
1.1 MBAdobe PDFView/Open Request a copy


Items in UCSC Digital Library are protected by copyright, with all rights reserved, unless otherwise indicated.