A compact feature set for recognition of handwritten numerals and vowels in the Kanarese script
R. Santhosh, Gowri Srinivasa · 2012
In this work we present a handwritten character recognition system for vowels and numerals of the Kanarese script. The system reads a handwritten character in the form of an image and translates it to a machine editable form. We have accomplished this through building a dataset of handwritten characters (collected digitally from a number of subjects who are versed with the Kanarese script). We have also designed a small set of features that discerns whether the input is a numeral or a vowel and thereafter classifies it. We demonstrate the efficacy of the features using a decision tree-based classifier. The feature set, which is compact, achieves a high average testing accuracy.