Designing for Accessibility
summary
A red hearing aid shifts stigma into belonging and motivates design-for-accessibility. A speech project called Project Euphonia trains models on atypical voices: one collaborator records ~15,000 sentences and a custom recognizer suddenly understands where the cloud model fails, proving per-person adaptation can work though generalization remains unsolved. A browser app called Teachable Machine (built on TensorFlow.js) lets users train local image classifiers for facial triggers to control smart-home actions as a rapid, on-device prototype.
AI-written summary · about
chapters (12)
- 0:00Start
- 4:30Momentary Disabilities
- 7:46Project Euphonium
- 9:08Google Cloud Speech Recognition
- 10:04Speech Recognizers
- 15:29Speaking Modalities
- 15:56Steve Saleen
- 16:44Gaze Pointer
- 17:43Machine Learning Tool
- 19:07Why Is this Useful
- 19:31Different Models for Different Use Cases
- 21:50Voice Samples
related talks
51mDisability Inclusion Beyond Accessibility Standards
13mBeyond Barriers: AI Reshaping Digital Accessibility
34mAI and Accessibility: the Good, the Bad, and the Bollocks
48mProgramming for accessibility
30mAccessibility sparks innovation: Democratize access to hearing devices
36mMaking Technology More Accessible with .NET and AI
30mAI and Accessibility: the Good, the Bad, and the Bollocks
41mProgramming for Accessibility