Eyes-free text entry with error correction on touchscreen mobile devices

Abstract

We present an eyes-free text entry method for mobile touchscreen devices. Input progresses by inking Graffiti strokes using a finger on a touchscreen. The system includes a word-level error correction algorithm. Auditory and tactile feedback guide eyes-free entry using speech and non-speech sounds, and by vibrations. In a study with 12 participants, three different feedback modes were tested. Entry speed, accuracy, and algorithm performance were compared between the three feedback modes. An overall entry speed of 10.0 wpm was found with a maximum rate of 21.5 wpm using a feedback mode that required a recognized stroke at the beginning of each word. Text was entered with an overall accuracy of 95.7%. The error correction algorithm performed well: 14.9% of entered text was corrected on average, representing a 70.3% decrease in errors compared to no algorithm. Where multiple candidates appeared, the intended word was 1 st or 2 nd in the list 94.2% of the time.

i'sFree

“…In addition to tap typing, some research investigated gesturebased text entry on touchscreen, such as an eyes-free Graffiti text input system [23,25] and a swipe-based invisible text entry method with numberpad-like layout on smartwatches [17].…”

Typing Performance of Blind Users

“…While the visual layout of the QWERTY keyboard is identical to that presented to sighted users, the text-entry rates are much slower for visually impaired users [19]. To address this problem a number of works have proposed novel interfaces for non-visual text-entry on mobile touchscreen devices; including new keyboard layouts [3,8] and alternative methods of inputting text [1,16,19,22,23].…”

Combining Auditory and Visual Menus

“…Text entry can also be implemented using ad i ff erent touch screen gesture for each character.T inwala and MacKenzie [24] used gestures that resemble letters as input and auditory and tactile feedback to guide eyes-free entry.Letters were entered one at atime and word-levelerror recognition with adictionary wasused to improve accuracy. Tinwala and MacKenzie suggested that changing the speech feedback from the character-leveltothe wordlevels peeds up writing and lessens user frustration.…”