Studying the variability of handwriting patterns using the Kinematic Theory.

The variability observed in handwriting patterns is analyzed from the perspective of integrating the resulting motor control knowledge in the design of more powerful handwriting recognizers in personal digital assistants (PDAs) and smartphones. Using the highest representational level of the Kinematic Theory of Rapid Human Movement, the Sigma-Lognormal model, this article reports basic theoretical and practical results that could be taken into account in the design of such systems. The main movement variability introduced by the neuromuscular system (NMS) and induced through the scheduling of motor tasks by the central nervous system (CNS) is divided into global and local fluctuations. From a fiducial action plan decoded by this model, a wide range of handwriting distortions are artificially generated by acting on the Sigma-Lognormal parameters. The resulting patterns are studied to understand scale changes and rotational deformations, the two basic features that a recognizer has to take into account. An experiment based on the writing of the same word by six writers is also reported. The results, obtained by an ANOVA analysis, corroborate the predictions and support the relevance of the Kinematic Theory for the analysis and synthesis of handwriting disruptions. These findings consolidate the results of previous studies on single strokes using the Sigma-Lognormal model. Overall, this report provides new insights into our understanding of motor control, as well as into practical cues for the development of huge databases of letters and words to train and test on-line handwriting classifiers and recognizers.