EYE-GAZE WORD-PROCESSING

Erica is a personal computer operated by eye-gaze. Simply by looking at menu options displayed at different locations on the computer monitor, a disabled user can invoke commands without the need for standard input devices. The design of text entry software for Erica’s-gaze word processor is described. Because gaze-point resolution is limited, the full-screen image of a keyboard cannot be employed. Instead, the design challenge is to find a functional and friendly encoding of text characters using individual menu screens with only six choices on each screen. A basic encoding is accomplished through a static tree-structured menu system, which associates the 85 most common ASCII characters with unique sequences of menus and eye-gaze picks. The expected length of the pick sequences is minimized based on the approximate frequencies of individual characters determined from a 55 882-character corpus of English language text. Text entry speeds are further enhanced by adding context-sensitive nodes to the static tree. The menu screens corresponding to these nodes contain the five characters most likely to follow the last two characters entered, based on character transition probabilities also derived from the sample corpus. The context-sensitive design is shown to decrease text-entry times by an average of 25% relative to the static tree without character prediction. © 1990 IEEE