The intelligent numeric keypad

The intelligent numpad attempts to address two issues: First, to provide a low-cost simple Braille chording keyboard, and second, to explore the possibility of automatically detecting the input mode from the temporal and spatial keyboard dynamics.

The intelligent numpad is realised using standard low-cost off-the-shelves numeric keypad. The particular model used in this experiment is a convenient wireless keypad. The figures below shows how the numpad can be used for regular digit input or cursor navigation, and as a Braille chording keyboard by using it upside down.

This implementation supports the English Braille alphabet. Note that only the basic letters are supported.

The following Figures show the Braille keypad is used in detail. The user can chose to use whichever keys for Braille-chords. The device will automatically detect the Braille-dot-to-key mapping.

Traditional numeric keypads can be used to input digits and for cursor navigation and the NumLock key is used to toggle between the two modes. The problem is that the user will not know which mode is active before observing the feedback from the system, i.e., the output produced. Some numeric keypads provide a LED that indicates mode, but does LED-on indicate digit-input mode or cursor navigation mode? The detection of mode is therefore heavily dependent on visual feedback.

Mode is automatically detected by investigating the spatial and temporal keystroke dynamics. Keystrokes involving multiple keys indicate chords and indicate that the device is in Braille input mode. Furthermore, Braille chords are input at slower regular patterns compared to digits, while digits are often input in a rhythmic pattern related to the cognitive chunking of the digits. Finally, cursor navigations can be identified by short burst of rapid and regular keystroke sequences, such as when the user detects a mistakes, and move the cursor backwards rapidly to make the correction.

The Braille chording keyboard is applicable to blind users familiar with Braille. The automatic detection of mode is more generally applicable. You can read more about the intelligent keypad in:

• Frode Eika Sandnes and Yo-Ping Huang, "The Intelligent Numpad: Detection of Mode from Spatial and Temporal Keystroke Characteristics", Proceedings of IEEE System, Man and Cybernetics IEEE SMC'06 Conference, Taipei, Taiwan, 8-11 Oct, IEEE Press, pp. 2446-2551, 2006.

A description of a language model for detecting and automatically correct errors for such input strategies is documented in:

• F. E. Sandnes, Y-P. Huang, “Chord level error correction for portable Braille devices”, IEE Electronics Letters, Vol. 42, No. 2, pp. 82-83, 2006.
• Frode Eika Sandnes, Yo-Ping. Huang, Non-Intrusive Error-Correction of Text Input Chords: A Language Model Approach, The 2005 North American Fuzzy Information Processing Society Annual Conference, Soft Computing for Real World Applications, Ann Arbor, Michigan, USA, IEEE Press, June 22–25, pp.373-378, 2005.

Early work addressing keystroke dynamics and the regularity of input is documented in:

• Frode Eika Sandnes and Hua-Li Jian, ”Pair-wise Variability Index: Evaluating the Cognitive Difficulty of using of Mobile Text Entry Systems”, Lecture Notes on Computer Science 3160, pp.347-350, 2004.