A study of calligraphic skill by virtual brush-writing with haptic device: Hidden Markov modeling of writing strokes

Handwritten characters involve much variant appearance by each individual. Especially, brush-writing, such in Japanese or Chinese calligraphy, is very skillful action, since the writer is required to activate the calligraphic technique, such as the pullout, the stop, and flick, with softness of the brush. In this article, we implemented a virtual brush-writing system with the pen-style haptic device which enables human-computer interaction through the touch with force-feedback. The system provides the reaction force against the virtual paper, and records the brush stroke by means of brush-tip position and pressure during the virtual writing. In order to describe the strokes by each writer, we utilized Hidden Markov model (HMM). It is regarded as that the observation sequence and hidden state, in terms of HMM, are corresponding to the stroke (pressure and velocity sequence) and the calligraphic technique. The model was trained by the stroke of six writers, and these models were experimentally verified for classification and evaluation of the similarity among the strokes of each writer. It might be considered that the likelihood through the HMM parameters suggests the measure of calligraphic skill of writer.