Improving Gesture Recognition Accuracy on Touch Screens for Users with Low Vision
summary
Compares five nearest-neighbor gesture recognizers on 2,400 samples from 10 users with low vision and 10 sighted users, finds baseline P recognizer drops from 95.9% to 83.8% for low-vision gestures, and implements two concrete fixes: one-to-many point alignments and a turning-angle–augmented point distance that produce a new P+ variant achieving 94.7% average accuracy (up to 98.2%) and about 3× faster matching than the original P.
AI-written summary · about
chapters (9)
- 0:00Introduction to research
- 2:35Study methodology and dataset
- 6:13Gesture consistency analysis
- 9:30Shape and bending error findings
- 10:57P dollar recognizer overview
- 12:21Modifications to P dollar
- 14:51Results of modified recognizer
- 16:18Conclusion and contributions
- 17:21Q&A session
related talks
14mSmart Touch: Improving Touch Accuracy for People with Motor Impairments with Template Matching
11mTilt-Explore: Making Tilt Gestures Usable for Low-Vision Smartphone Users
20mScene text access: a comparison of mobile OCR modalities for blind users
06mRecognizing Unintentional Touch on Interactive Tabletop
21mVizLens: A Robust and Interactive Screen Reader for Interfaces in the Real World
21mA Face Recognition Application for People with Visual Impairments: Understanding Use Beyond the Lab
23mRealPen: Providing Realism in Handwriting Tasks on Touch Surfaces using Auditory-Tactile Feedback
58mAccessible Spatial Computing