User Preference Learning System for Tangible User Interfaces

There is increasing demand for tangible user interfaces(TUIs), for easy and intuitive control of applications. In order to allow use of TUI devices for controlling applications,a system needs to associate functions with the TUI devices. Although several TUI applications and toolkit shave been proposed, it is difficult to use them for various applications since they require the user to perform several inconvenient operations. Therefore, we propose a preference learning algorithm for an automatic matching between TUI devices and application functions. A user can freely place and easily operate the TUI devices, and our system creates and learns user's preference data. Our system does this by determining three concrete types of user preferences specialized for each application and six abstract types of those common to all applications. Our proposed system has achieved over 80% accuracy in automatic matching.