Enabling Auto-Correction on Soft Braille Keyboard

Area of Science:

• Assistive Technology
• Human-Computer Interaction
• Computational Linguistics

Background:

• Soft Braille keyboards are crucial for visually impaired individuals on smartphones.
• Current Braille keyboards face significant accuracy and efficiency challenges.

Purpose of the Study:

• To develop an intelligent Braille keyboard with auto-correction capabilities.
• To enhance text input accuracy and efficiency for visually impaired users on mobile devices.

Main Methods:

• Utilized optimal transportation theory to calculate distances between touch inputs and Braille patterns.
• Integrated a language model to predict word probabilities for auto-correction.
• Evaluated the system through touch interaction simulations and user studies.

Main Results:

• Reduced error rates from over 55% to 19.80% in simulations under high typing noise.
• User studies showed a 59.5% reduction in word error rate (WER) for blind participants.
• Achieved a slight decrease in words per minute (WPM) with significantly improved accuracy.

Conclusions:

• The intelligent Braille keyboard demonstrates superior performance over existing solutions.
• This approach has the potential to greatly enhance the typing experience for Braille users on touchscreen devices.
• Auto-correction is a key factor in improving usability for assistive text input methods.