• DocumentCode
    425380
  • Title

    Epipolar Contrained User Pushbutton Selection in Projected Interfaces

  • Author

    Kale, Amit ; Kwan, Kenneth ; Jaynes, Christopher

  • Author_Institution
    UK Center for Visualization and Virtual Reality, Lexington KY
  • fYear
    2004
  • fDate
    27-02 June 2004
  • Firstpage
    156
  • Lastpage
    156
  • Abstract
    An almost ubiquitous user interaction in most HCI applications is the task of selecting one of out of a given list of options. For example, in common desktop environments, the user moves the mouse pointer to the desired option and clicks it. The analog of this action in projector-camera HCI environments involves the user raising her finger to touch one of the different virtual buttons projected on a display surface. In this paper, we discuss some of the challenges involved in tracking and recognizing this task in an projected immersive environment and present a hierarchical vision based approach to detect intuitive gesture-based "mouse clicks" in a front-projected virtual interface. Given the difficulty of tracking user gestures directly in a projected environment, our approach first tracks shadows cast on the display by the user and exploits the multi-view geometry of the camera-projector pair to constrain a subsequent search for the users hand position in the scene. The method only requires a simple setup step in which the projector\´s epipole in the camera’s frame is estimated. We demonstrate how this approach is capable of detecting a contact event as a user interacts with a virtual pushbutton display. Results demonstrate that camera-based monitoring of user gesture is feasible even under difficult conditions in which the user is illluminated by changing and saturated colors.
  • Keywords
    Application software; Computer displays; Computer vision; Event detection; Fingers; Geometry; Human computer interaction; Mice; Virtual reality; Visualization;
  • fLanguage
    English
  • Publisher
    ieee
  • Conference_Titel
    Computer Vision and Pattern Recognition Workshop, 2004. CVPRW '04. Conference on
  • Type

    conf

  • DOI
    10.1109/CVPR.2004.58
  • Filename
    1384954