Hand-written Urdu Numerals Recognition Using Kohonen Self Organizing Maps

Abstract

Urdu is one of the widely spoken languages in the world. This paper presents utilization of Kohonen Self Organization Maps for hand-written Urdu numerals. A dataset of hand-written Urdu numeral, comprises 6200 numerals, was collected from undergraduate students. Half of this data set was used for training a 10 by 10 Kohenen Self Organizing Maps. The trained clusters of the Self Organizing Maps were assigned with labels manually based on the visual appearance. The whole data set was then used for testing the trained Neural Network. Results demonstrated promising performance of the proposed scheme.