Effect of variability on letters generation with the vectorial delta-lognormal model

Abstract

One of the primary goals of handwriting modeling is to understand how humans represent, control and generate compex movements. Moreover handwriting modeling has been also used for practical applications such as handwriting analysis and recognition. Numerous models used to date were not strong enough to explain and support some fundamental results about biomechanical or neurophysiological systems and neither practical enough to be used for accurate handwriting generation in the kinematic and the spatial domains. The vectorial delta lognormal model has been shown, in the past few years, to answer to these two paradigms showing accuracy for simple movements simulation and flexibility for letters and cursive handwriting generation. In this paper we show how this model can help to study handwriting variability, particulartly, the effects of the fluctuations of the commands and of the neuromuscular effectors, on the movement generated. Some characters models are proposed with examples of the variability effects. A parametric representation of allographs can then be used to represent basic shapes and some models of distortion, to generate a variety of prototypes.