Group Touch: Distinguishing Tabletop Users in Group Settings via Statistical Modeling of Touch Pairs
summary
Uses only built‑in FTIR touch data from Microsoft PixelSense, modeling distance, orientation, time between touch pairs with an MLP and a greedy grouping algorithm to decide if two touches come from the same person. Validated on 17 classroom sessions with high‑school students, it achieves 92.92% mean accuracy (SD=3.94%) using a 0.8 probability threshold. Produces short‑term, non‑persistent user groups, works well for gesture disambiguation and conflict detection, and fails mainly when inter‑touch gaps exceed ~2.5 minutes.
AI-written summary · about
chapters (10)
- 0:00Introduction to Group Touch
- 2:11Challenges in Classroom Settings
- 5:00Group Touch Approach Overview
- 8:09Study Design and Data Collection
- 9:15Model Features and Pairing Touches
- 11:02Grouping Algorithm and Threshold
- 12:47Evaluation and Accuracy Results
- 15:19Use Cases and Applications
- 16:45Future Work and Limitations
- 17:07Conclusion and Q&A
related talks
16mInvestigating the feasibility of finger identification on capacitive touchscreens using deep ...
17mPredicTouch: A System to Reduce Touchscreen Latency using Neural Networks and Inertial ...
37mMulti-touch Attribution: What am I training for?
18mInfiniTouch: Finger-Aware Interaction on Fully Touch Sensitive Smartphones
12mEclipseTouch: Touch Segmentation on Ad Hoc Surfaces using Worn Infrared Shadow Casting
18mEstimating the Finger Orientation on Capacitive Touchscreens Using Convolutional Neural Networks
14mLeveraging Dual-Observable Input for Fine-Grained Thumb Interaction Using Forearm EMG
19mATK: Enabling Ten-Finger Freehand Typing in Air Based on 3D Hand Tracking Data