The problem of extraction, description, and estimation ofhandwriting deformation in online character recognition is discussed.First, feature-point correspondence is extracted between an inputpattern and a reference pattern by elastic matching and a deformationvector field (DVF) is generated. Second, the DVF is expanded into aninfinite series by applying iterative local affine transformations.Finally, the interpattern distance is calculated between the inputpattern and the reference pattern superposed by low-order components oflocal affine transformations. Recognition tests made on cursive kanjicharacter data have revealed the high discrimination ability of thismethod
展开▼