Handwritten digits recognition based on swarm optimization methods

Abstract

In this paper, the problem of handwritten digits recognition is addressed using swarm based optimization methods. These latter have been shown to be useful for a wide range of applications such as functional optimization. The proposed work places specific swarm based optimization methods that are the particle swarm optimizer and variations of the bees' colony optimization in handwritten Arabic numerals recognition so that to improve the generalization ability of a recognition system through the use of two alternatives. In the first one, swarm based methods have been used as statistical classifiers whereas in the second one a combination of the famous gradient descent back-propagation learning method and the bees algorithm has been proposed to allow better accuracy and speediness. Comparative study on a variety of handwritten digits has shown that high recognition rates (99.82%) have been obtained. © 2010 Springer-Verlag.