Investigating the feasibility of finger identification on capacitive touchscreens using deep ...
summary
Study collects a representative dataset of 455,709 capacitive images from a Nexus 5 (27×15 pixel touch maps), trains CNNs on raw pixels and achieves a position‑invariant left vs right thumb accuracy of 93% (validation 92%). Per‑tap classification runs at 19.2 ms on-device. Two sample apps—a porous dual‑app UI and a painting app with thumb shortcuts—are implemented and evaluated; effective user‑level accuracy in the porous UI is 85.2% and participants report the interaction as useful and learnable.
AI-written summary · about
chapters (12)
- 0:00Introduction and motivation
- 1:55Previous approaches and limitations
- 3:40Research goal and approach
- 5:06Data collection study
- 6:32Feature exploration and modeling
- 7:56Model improvements and results
- 9:00Model architecture and deployment
- 10:05Use case demonstrations
- 11:13User study and evaluation
- 13:21Results and user feedback
- 13:42Conclusion and future work
- 14:48Q&A session
related talks
18mInfiniTouch: Finger-Aware Interaction on Fully Touch Sensitive Smartphones
18mEstimating the Finger Orientation on Capacitive Touchscreens Using Convolutional Neural Networks
14mDeepFisheye: Near-Surface Multi-Finger Tracking Technology Using Fisheye Camera
14mLeveraging Dual-Observable Input for Fine-Grained Thumb Interaction Using Forearm EMG
17mPredicTouch: A System to Reduce Touchscreen Latency using Neural Networks and Inertial ...
19mGroup Touch: Distinguishing Tabletop Users in Group Settings via Statistical Modeling of Touch Pairs
19mATK: Enabling Ten-Finger Freehand Typing in Air Based on 3D Hand Tracking Data
10mHandyTrak: Recognizing the Holding Hand on a Commodity Smartphone from Body Silhouette Images