Dynamic Bayesian Adjustment of Dwell Time for Faster Eye Typing

Area of Science:

• Human-Computer Interaction
• Assistive Technology
• Biomedical Engineering

Background:

• Eye typing offers hands-free computer interaction, crucial for individuals with upper limb disabilities.
• Current eye typing methods use fixed dwell times, creating a speed-accuracy trade-off.
• Longer dwell times reduce errors but slow down input, while shorter times increase errors.

Purpose of the Study:

• To enhance eye typing speed and accuracy for hands-free human-computer interaction.
• To introduce a novel method for dynamically adjusting dwell times based on input history.
• To maintain low error rates while significantly increasing typing speed.

Main Methods:

• Developed a probabilistic generative model of gaze to predict likely characters.
• Dynamically adjusted dwell times for each key based on past typing history.
• Evaluated the model on both able-bodied participants and individuals with spinal cord injury (SCI).

Main Results:

• Achieved significant increases in typing speed for both able-bodied (41.8%) and SCI (49.5%) subjects.
• The dynamic dwell time method maintained low error rates compared to standard fixed dwell times.
• Observed greater inter-subject variability in typing performance among SCI participants.

Conclusions:

• Dynamic dwell time adjustment based on gaze prediction is an effective method to improve eye typing efficiency.
• This approach offers a promising solution for faster and more accurate hands-free computer access for users with disabilities.
• Further research may explore personalized models to account for inter-subject variability, particularly in SCI populations.