|
Targets, in a vision-based user interface, are similar to buttons in a
traditional graphical user interface. They are a region in space
that the user can touch to perform an action.
The simplest targets can trigger binary events. They are trained to recognize an object in a region of an image by its color pattern. They can events when the object first appears, and when it disappears. Technically, they match the color histogram of a region of an image with the histogram obtained during training. Since two histograms will never match exactly, there is a threshold on the similarity of the two histograms. The amount of variation which the target can tolerate in the object's appearance depends on how that threshold is set. We expose the threshold to the user in the form of a "sensitivity" parameter. 
More complex targets can respond more subtly to the user's touch. These provide continuous parameters describing how the user is touching the target. The quality of touch can map to control functions in various ways. One approach is similar to the musical concept of attack / sustain / decay for notes, which make them well suited to use for musical performance. Targets are trained by the user identifying a region in the image, typically a rectangle that can be moved and resized. The user then demonstrates the action they want the target to recognize. Options are available to allow the user to configure the events generated by the target. This seemingly simple interaction paradigm can be applied in numerous ways. In TouchFree Switch a single target is used to control scanning software. We have also used them as the mouse button to work in conjunction with buttonless cursor moving tools, like the Origin head tracker. We have used them to move the cursor by having two targets, one to control vertical movement, one to control horizontal movement. We have even created an "Air Piano" by placing a line of targets across the top of the image. |
| Contact: Rick Kjeldsen | Last updated: 6/7/02 | ||
 |
 |
 |
 |