Markus Kohler, Sven Schröter, Heinrich Müller Informatik VII (Computer Graphics) University of Dortmund, D-44221 Dortmund, Germany
The idea of computer vision-based interaction is that a computer observes the user with video-cameras, and interprets his actions using techniques of image recognition. In the past few years, vision-based interaction has found significant research interest, and about 40 projects and systems have come to our knowledge, which emphasize different applications and technical solutions ( Kohler 1998). First devices based on this technology did appear commercially ( Siemens 1998).
In the very general scenario of this contribution, several box-shaped sensitive regions are defined in a room. The user may select one of those regions by pointing towards it, in order to interact with a process attached to the region by hand gesture input and visual and auditive feedback. If the dialog is terminated, a selected region may also be deactivated. A concrete application which is the background of our development of this generic concept is remote control of home devices linked together by a domestic network. The idea in particular is to provide elderly and disabled users with a natural and intuitive means of controlling home appliances. In that application, the sensitive regions correspond to the different home devices which can be selected and manipulated by pointing at them (figure 1).
In order to implement this scenario, we have developed the ARGUS architecture outlined in figure 2. The system is roughly split in two parts, image processing depicted by the large block in the lower part of the figure, and global environment control represented by the smaller block at the upper part of the picture. The image processing part itself is subdivided in low level image processing and high level image processing. The following sections give a closer look at them. More details beyond those presented in this contribution, and references to literature concerning the applied methods can be found elsewhere ( Kohler 1999).