To double the effective rate of communications for eye gaze communicators from an average of 12 words per minute to 24 words per minute
Create an open source research platform & publish results of experiments
Quantitatively validate concepts through metrics enabled experimentation
Perform experiments with known best practices of AAC design
Combine the results into an open reference platform for eye gaze speech
Initially scoped to English language speakers
- Lower letter selection error rates through evaluating and selecting best eye gaze sensor technology
- Lower letter selection error rates through adaptive noise suppression (e.g. dynamically adjusting noise suppression based on observing signal jitter)
- Lower letter selection error rates through language prediction models that apply the probability of letter selection to keyboard signal
- Word prediction / expansion using personalized dictionaries
- Word expansion using language stemming selections
- Word prediction using environment context
- Word prediction using sentence context
- Phrase prediction using environment & paragraph context & personal history
- Contextual overlay of predictions within the eye’s foveated cone
- Eye gaze enabled editing & correction
The contents of the repository are licensed under the MIT license.
Data is contributed to the dataset owned by Team Gleason under the Speak Faster Data Contribution Agreement license.
Data is sub-licensed to researchers under the Open Dataset License with De-Identification Restrictions.