You will need a one-way mirror for this. It's related to something called front projection.
This setup is typically, except from the touch part, used with auto-cues (text projected in front of a camera that f.ex. a news anchor reads from appearing looking into the camera).
For example:

For a full body you would of course need a bigger mirror and you would need to project the image from top (or side), turn the mirror 90 degrees so it reflect what comes from top (or side), as well as a layer between the projector and mirror (can be as simple as a cloth)

In addition to that, you would need a video mixer so you can super-impose/mix in the interface that is supposedly touched (same as the projected image from the TV or projector) (alt. do it in post with composition/video editing software).
The tricky part is to get the overlayed video (the user-interface) synchronized with the touches from the talent. You would need initially to perform some calibration, f.ex. an image with fixed corners and a center that is projected while the talent touches these spots. Then scale/zoom/place video/camera to match the super-imposed image that is projected.
The other tricky part is the lighting conditions. The room need to be relative dark for the mirror to work properly at the same time as the talent is properly lit for the camera. You can darken the area behind and around the mirror with black colored textile in addition to turn off all none-talent lights and so forth.
Hope this helps!