|A relative keyboard, or predictive keyboard is a textual input method whereby
'keystrokes' are detected via machine vision, as movements relative to one another,
instead of a discrete, unambiguous sequence of key-presses activating electrical
If the user is a touch-typist, never needing to look at the keys, this means
they can type on any haptic-enabled surface, without requiring visual keys,
or indeed any form of key at all.
The first prototype, created by two researchers at the Language Technologies
Institute of Carnegie Mellon University in the US, during late 2007-early 2008,
the first relative keyboard design is a blank touchscreen, allowing multiple
simultaneous touches, and simply working out what key is desired by the co-ordinates
pressed on screen.
Their software works out what is being typed by measuring the relative distance
between keystrokes. Filtering possible strings using a dictionary makes it possible
to work out what word was meant, and so reduce error rates - as going back over
your own work to correct typos without a visual frame of reference would be
an utter nightmare.