Precise hand-printed character recognition using elastic models via nonlinear transformation

Distorted character recognition is a difficult but inevitable problem in hand-printed character recognition. In this paper, we propose a character recognition method using elastic models for recognizing cursive characters with intricate structure. The models are fitted to unknown input patterns by applying the EM algorithm to minimize a measure of fitness. To avoid falling into local minima, multiresolution approach is introduced. Moreover, nonlinear transformation is adopted to realize more flexible matching. Experiments performed on Japanese characters show effectiveness of the proposed method