A structural point feature based Neural Network Approach for Handwritten word Recognition

Abstract

Character recognition is one of common but major application of image processing. It is used to convert published and handwritten documents in digital form. The character recognition has various complexities in terms of different fonts, alignment, strokes. With the inclusion of words and lines the complexities also increase.  In this paper, the structural point information is extracted from hand written words. A layer based work model is designed in this research. As the work is applied on real-time characters, in the first layer the nosie is removed by using Gaussian filter. In the second stage, the structural points of image are extracted by using mathematical filter. In the final layer, the neural network is applied for classifying and recognizing the words. The results identified that the proposed model achieved the effective accuracy.