next up previous
Next: Conclusions on the Up: The `visual memory Previous: A wearable face-recognizer

Dual adaptation space

Using the `reality mediator', I repeated the classic experiments like those of Stratton and Anstis (e.g. living in an upside-down or negated world), as well as some new experiments, such as learning to live in a world rotated 90 degrees.

Of note, however, I observed that visual filters differing slightly from the identity (e.g. rotation by a few degrees) had a more lasting impression on me when I removed my apparatus (e.g. left me incapacitated for a greater time period upon removal of the apparatus), than visual filters that were far from the identity (e.g. rotation by 180 degrees -- upside-down). Furthermore, the visual filters close to the identity tended to leave me with an opposite aftereffect (e.g. I'd consistently reach too high after taking off the RM where the images had been translated down slightly, or reach too far `clockwise' after removing the RM that had been rotating images a few degrees counterclockwise). Visual filters far from the identity (such as reversal or upside-down mappings) did not leave me with an opposite aftereffect: I would not see the world as being upside down upon removing upside-down glasses. I think of this phenomenon as being analogous to learning a second language (either a natural language or computer language). When the second language is similar to the one we already know, we make more mistakes switching back and forth than when the two are distinct.

The `visual memory prosthetic' was based on a partially mediated reality[2], that is, only part of the visual field of view was mediated, in this case, with the computer-induced (and sometimes annotated) flashbacks.

The reason for using the `rot90' (rotate 90 degrees) arrangement was twofold: firstly this matched the aspect ratio of the face (which is generally taller than it is wide, in fact it is generally about 4 units high and 3 wide which exactly matches `rot90' video), and this created a distinct dual adaptation space. At one time I tried also rotating the display so that it would match the rotated camera, but later went back to having the camera in `portrait' mode and the display in `landscape' mode.

When two (or more) adaptation spaces were distinct, for example, in the case of the identity map (unmediated zones of the glasses) and the rotation operation (`rot 90'), I could sustain a dual adaptation space and switch back and forth between the `portrait' orientation of the identity operator and and `landscape' orientation of the `rot 90' operator without one causing lasting aftereffects in the other.

Often the mediated and unmediated zones are in poor register and I cannot fuse them. The poor register may even be deliberate, e.g. I often like to have my right eye in a rotated (`rot 90') world even though this means that I cannot see in stereo in a meaningful way. However, I can still switch my concentration back and forth. I am able to selectively decide to concentrate on one or the other of these two worlds.

Again, I can switch my attention back and forth between the mediated reality and ordinary vision. I see a double-vision effect (e.g. when I look at someone's face through the glasses of Fig 4, I often see two replicas of their face, the one that is mediated, and the one that is not). This doubling effect, due to imperfect registration between the mediated and unmediated zones, may or may not be a problem depending on how the RM is used. For example, if I present the mediated world as grey, it remains distinct from the unmediated world, and I am able to mentally switch back and forth between seeing directly, and living in the mediated world, even though the two overlap almost exactly. However the `rot90' was found to provide the best distinction between the unmediated and mediated realities.

next up previous
Next: Conclusions on the Up: The `visual memory Previous: A wearable face-recognizer

Steve Mann
Wed Feb 14 01:19:59 EST 1996