Conceptual-level-learning-based Refinement of a Human-motion Model
Hirohide Ushida, A. Imura, Toru Yamaguchi, Tomohiro Takagi · IEEJ Transactions on Electronics Information and Systems · 1995
This paper introduces conceptual level learning (CLL) as a method that can be used to refine human-motion models. CLL imitates human learning processes in which people learn through multi-modal macro-instructions represented by language, gestures, and other means. Our method implements a human-motion model and a macro-instruction interpreter using fuzzy associative memory systems. The model has several nodes, and their connections represent the relationship between upper concepts and minimum components in terms of motion. The interpreter also has several nodes, and their connections represent the relationship between macro-instructions and trends in the minimum components. The interpreter transforms multi-modal macro-instructions to minimum components. Our method provides unskilled people with an easy-to-use design system and provides for useful intelligent human-machine interaction such as virtual reality. This is done by real-time refinement of a model, and visualizing motion so the user can observe it. Here, this method is used to construct and refine a human tennis-motion model.