Ensembles of classifiers for handwritten word recognition specialized on individual handwriting style

Abstract

The study of multiple classifier systems has become an area of intensive research in pattern recognition recently. Also in handwriting recognition, systems combining several classifiers have been investigated. Recently, new methods for the generation of multiple classifier systems, called ensemble methods, have been proposed in the field of machine learning, which generate an ensemble of classifiers from a single classifier automatically. In this paper a new ensemble method is proposed. It is characterized by training each individual classifier on a particular writing style. The new ensemble method is tested on two large scale handwritten word recognition tasks. © Springer-Verlag 2004.