Classification of Typed Characters Using Backpropagation Neural Network

This thesis concentrates on classification of typed characters using a neural network. Recognition of typed or printed characters using intelligent methods like neural network has found much application in the recent decades. The ability of moment invariants to represent characters independent of position, size and orientation have caused them to be proposed as pattern sensitive features in classification and recognition of these characters. In this research, uppercase English characters is represented by invariant features derived using functions of regular moments, namely Hu invariants. Moments up to the third order have been used for the recognition of these typed characters. A single layer perceptron artificial neural network trained by the backpropagation algorithm is used to classify these characters into their respective categories. Experimental study conducted with three different fonts commonly used in word processing applications shows good classification results. Some suggestions for further work in this area have also been presented.

Text FK_2001_2.pdf Download (2MB)

Actions (login required)

View Item