MagicSuit
> MagicSuit is a new type of virtual reality application where all user interaction is handled through a motion capture suit. Gesture recognition methods used to make the virtual environment respond as if the user possessed magical powers. In addition, MagicSuit features a powerful physics engine and advanced graphics techniques designed to enhance the realism of the virtual environment. The player can interact with the world in a variety of ways, including directing a flying cloud of magic dust, creating fireballs out of thin air, and commanding normally inanimate objects to fly to different locations. MagicSuit has been the subject of two of my papers, neither of which have yet been published.
> Gesture recognition is the power behind MagicSuit. When the player wears a motion capture suit, the positions of his limbs are recorded and his pose, or gesture, is compared to predefined gestures called targets. Two types of targets are used in MagicSuit: PCA(Principle Component Analysis) targets use statical pattern-matching to turn a particular gesture into a position in space. This point is compared to the point occupied by a target gesture, and if the distance is smaller than a certain threshold, a match is declared. The other method, called discrete feature analysis, looks at specific parts of the user's pose, such as the angle of the elbow or the height of the hand, to determine if the conditions for a particular gesture (and its response) are met. >
> MagicSuit serves as an excellent testbed for PCA-based gesture recognition algorithms. Icons located on one wall of the virtual environment represented the user's "distance" from predefined target poses. Test subjects were asked to perform 32 gestures, and the results were recorded and analyzed. The PCA system proved to be very good at identifying gestures and poses performed by the subjects, regardless of their size or shape. >