Everyone Can Do Magic: An Interactive Game with Speech and Gesture Recognition

摘要

This paper presents a novel game design that allows players to learn how to cast magic spells that combine hand gestures and speech. This game uses the imperfect recognition performance in speech and gesture recognition systems to its advantage to make the game challenging and interesting. Our game uses a Wii remote encased in a wand and a microphone to track player's gestures and speech which are then recognized to determine if they have performed the spell correctly. Visual feedback then provides confirmation of success. Through the game, players learn to adjust their speaking and movement patterns in order to meet the requirements of the recognition systems. This effectively mimics the characteristics of casting spells correctly such that players are trying to adjust their performance so that an "oracle" recognizes their speech and movement to have a magical outcome. A user study has confirmed the validity of the idea and establishes the accuracy required to create an interesting game based on the theory of channels of flow.