ATK: Enabling Ten-Finger Freehand Typing in Air Based on 3D Hand Tracking Data
summary
A system called ATK uses 3D hand tracking and a two-stage pipeline — deterministic tap detection (200 ms window, peak-velocity threshold alpha=0.45) followed by an augmented Bayesian classifier that combines a language model, a 3D spatial model and a finger-tapping model — to disambiguate taps and tapping finger. Motion-capture evaluation with eight participants yields 96.7% recall and precision 97% for tap detection, and text entry rises from 23.0 WPM to 29.2 WPM with near-zero corrected errors while using audio cues and thumb gestures for candidate selection.
AI-written summary · about
chapters (8)
- 0:00Introduction to air typing
- 2:00Relevant work review
- 3:45System design and methodology
- 5:57Initial study results
- 8:34Algorithm development
- 12:25Evaluation and user study
- 14:27Conclusion and future work
- 16:39Q&A session
related talks
20mTipText: Eyes-Free Text Entry on a Fingertip Keyboard
12mTapBoard 2: Simple and Effective Touchpad-like Interaction on a Multi-Touch Surface Keyboard
15mTip-Tap: Battery-free Discrete 2D Fingertip Input
18mInfiniTouch: Finger-Aware Interaction on Fully Touch Sensitive Smartphones
21mDualKey: Miniature Screen Text Entry via Finger Identification
14mSmart Touch: Improving Touch Accuracy for People with Motor Impairments with Template Matching
17mPredicTouch: A System to Reduce Touchscreen Latency using Neural Networks and Inertial ...
16mInvestigating the feasibility of finger identification on capacitive touchscreens using deep ...