Robotic Chinese Calligrapher: Action, Imitation and Learning

Chinese calligraphy is a unique form of traditional art combining function and aesthetics. It is spatial and temporal, with the characteristics of both painting and dancing. Visual entertainment can be derived from the 2- D calligraphy on the page (mainly calligraphic styles), and from the 6-D motion of the brush as it is written.;Due to the improvements in information technology, typing is gradually taking the place of handwriting. Transferring more calligraphic styles into computer fonts will both enrich the pool of available computer fonts and allow us to continue to appreciate Chinese calligraphy. Unfortunately, it is highly unlikely for a past calligrapher to have written all of the characters in his lifetime. This dissertation aims to automatically synthesize the unwritten characters while preserving the calligrapher's particular style. A calligraphy robot to write these characters is also developed. The robotic calligrapher also provides visual entertainment as it works. It is impossible for a student to determine the 6-D motion of the brush if he has no experience of writing calligraphy. The robotic calligrapher can also act as a teacher, demonstrating to students how to correctly manipulate the brush.;It this dissertation, the robotic calligrapher Callibot is introduced. Callibot can write characters in a particular calligrapher's style, even if this calligrapher did not write the particular characters in his lifetime. The contributions of this dissertation are as follows: 1) A new calligraphy robot, Callibot, is designed and developed. Callibot features analytic solutions to inverse kinematics, few singularities and large workspace. 2) A new method for transferring calligraphy skills from a human to the robot is proposed, the core of which is a "Learning from Demonstration" approach. After several demonstrations, Callibot can learn to write calligraphy as naturally as a human. This method is compatible with any robot with six or more degrees of freedom. 3) The generalized basic strokes of Chinese characters are newly defined. A new geometric approach to extract the basic strokes is proposed. If the stroke extraction approach is combined with the above learning approach, then Callibot can imitate calligraphy from images. 4) A new method to parameterize Chinese calligraphy is proposed. The similarity (distance) between styles is measured using the parameterization results, which are named Eigengraphies and Fishergraphies. And various classifiers can precisely classify calligraphic types and styles. 5) Anew method for synthesizing calligraphy in a given style with the help of the above similarity measuring approach is proposed. If the synthesis approach is combined with the imitation process, then Callibot can write the synthesized characters.