Character Modeling and Synthesis from Human Handwriting Motions with Esthetic Evaluation

Character Modeling and Synthesis from Human Handwriting Motions with Esthetic Evaluation

  We develop a scheme for constructing character font models from human handwriting data. First, handwriting motions are measured using a pen-tablet device, and stored as a 3-dimensional time series data. Then, the data is used to construct characters by employing B-splines and the theory of smoothing splines based on the minimum hand jerk model. We can then synthesize arbitrary words with arbitrary cursiveness. Moreover, by employing the concept of Induction Field in Vision, we propose a tuning method of the so-called smoothing parameter from an esthetic viewpoint of the resulting cursive characters. The effectiveness of the developed method is verified by some experiments, including those on calligraphic learning.