Encoding patterns for efficient classification by nearest neighbor classifiers and neural networks with application to handwritten Hindi numeral recognition

Abstract

Encoding of relevant information from visual patterns represents an important challenging component of pattern recognition. This paper proposes a contour-following based algorithm for extracting features from patterns. For classification of the encoded patterns by nearest neighbor (NN) classifiers, an iterative clustering algorithm is proposed to obtain a reduced, but efficient, number of prototypes. The algorithm works in a supervised mode and can perform cluster merging and cancelling. Moreover, mapping this NN classifier to a multilayer feedforward neural network is investigated. The performance of the algorithms is demonstrated through application to the task of handwritten Hindi numeral recognition. Experiments reveal the advantages of handling flexible sizes, orientations and variations.