International Conference on Advances In Engineering And Technology - ICAET 2014
Author(s) : MAMTA GARG , MUNISH KUMAR, SILKY BANSAL
Handwritten word recognition has been become a area of research in the field of pattern recognition. There is a rich literature available on word recognition in non-Indian scripts but limited work is available on Indian scripts. In this paper, we have presented an approach to recognize handwritten city names written in Gurmukhi script for postal automation. For recognizing words we have used a tree-diagonal feature extraction technique with SVM and k-NN classifiers. In this work, we have collected 18,000 samples of handwritten city names in Gurmukhi script. These samples have been collected from 60 different writers and each writer has written 30 city names. Using this approach, we have achieved a recognition accuracy of 90.8%.