Some observations on lognormality and motor control in handwriting

Lognormality has proven to be an effective way for handwriting modeling. It assumes that handwriting is a time superimposition of a sequence of commands issued by the central nervous system, each command producing a stroke, i.e. a movement with a lognormal velocity profile. Motor control theories, however, suggest that handwriting movements result from both central and local control, thus assuming that some movements of the sequence may not be the results of an explicit command issued by the central nervous system. In the light of those observations, we present an algorithm for segmenting handwriting movements into strokes, each of which corresponds to a command issued by the central nervous system, while disregarding those that may depend on local control. Experiments on handwriting samples show that the proposed algorithm detects the same number of strokes across multiple executions of an handwriting task by the same subject, and this set of strokes provides also a good reconstruction of the movement.