Recognition of Online Handwritten Gurmukhi Characters Through Neural Networks

Abstract

This paper recognizes online handwritten Gurmukhi characters and words. The neural network-based recognition for online handwritten Gurmukhi characters and words has been observed first time in this study. In this work, a scheme is proposed to develop a feature vector and its use as an input to neural network recognition engine. A set of low-level, high-level, and Gabor features are extracted, and a feed-forward neural network is trained to recognize 40 classes of Gurmukhi characters. This work implements rearrangement of strokes stage after recognition and post-processing stages. The results have been achieved as 93.53% and 83.69% for 4511 Gurmukhi characters and 2576 Gurmukhi words, respectively.