A distortion model of handwritten characters and evaluation of its power of expression

This paper discusses a distortion model for handwritten characters and its ability of representation. The following three assumptions are made in constructing the distortion model: (1) a standard figure is pre-set in correspondence to each character category; (2) the handwriting process is a sequential system with the standard figure as the input and the figure with distortion as the output; (3) the whole distortion is a combination of the deterministic and the stochastic distortions. The distortion model for the handwritten characters is constructed under the above assumptions. The feedback by the visual system during the handwriting process is regarded as a resorting force which affects a control on the stochastic distortion. The standard figure is considered as a combination of several elements, and the elements are estimated for each individual from handwritten character data. The distorted figure which is generated from the estimated standard figure, is considered to represent the handwritten character of the individual. Finally, the distance between figures is defined, and using it as the measure, the representation ability of the proposed distortion model is evaluated, indicating the usefulness of the model.