ANALYSIS OF HANDWRITING BY THE DYNAMIC-PROGRAMMING MATCHING

We examined the temporal invariance hypothesis in handwriting, using t he dynamic programming (DP) matching algorithm. The DP matching algori thm was originally developed for the speech recognition to eliminate f luctuations of time axis caused by different speech rates. This algori thm can determine the optimal point-to-point correspondences (warping function) between the standard velocity pattern and different patterns . We found that this algorithm worked well for various velocity patter ns, even when their total writing time was different from the standard pattern. Only 30 to 50 percents of handwriting movements showed tempo ral invariance. The writing movements which didn't show temporal invar iance consisted of several sub-motor units. The borders between the un its were not limited near the end of each letter. The results suggest that handwriting of a word does not show the temporal invariance and t hat it is executed by some discrete sub-motor units.