Handprinted numerals recognition by learning distance function

In designing more accurate character recognition, revealing the differences with other categories in distance function is important. In this paper, I propose Learning by Discriminant Analysis (LDA) as a method to learn distance functions. With a weigted Euclidean distance and a quadratic discriminant function as the original distance functions, LDA learns parameters by superposing the decision function for searching on the pattern set of the noticed category. The results for handwritten numeral recognition rate improved dramatically and its effectiveness was verified. In addition, when the values of the parameters after learning are changed and applied in a weighted Euclidean distance so that the misread patterns before learning are efficiently segmented and strong correlations exist between features, appropriate category boundaries are obtained