Intelligent Feature Analysis in Fuzzy Problem Domain: A Better Approach

The motivation of this paper comes from the need of generating feature vectors for classical pattern recognition problem. Though the paper basically deals with the problem of printed as well as handwritten character recognition, the idea of feature extraction and analysis may be successfully incorporated in other pattern recognition problem domains. To cope with the fuzzy ness of the problem space, some intelligent approaches have been presented in this paper. The features are fed to a classification mechanism that employs feed forward based recognition system, but instead of using general back propagation algorithm, resilient back propagation algorithm has been used by the classification system to achieve faster recognition rate and better accuracy. Empirical results of the system has been provided which shows that the feature extraction scheme presented in the paper is effective to meet the challenge of overcoming fuzzy ness of the problem domain