Swift : Turning Your Hand into a Writing Pad with Unmodified Smartwatches
Smartwatches remain constrained by small touchscreens. To enable expressive text input without extra hardware, we introduce Swift , an on-skin writing technique that treats the back of the hand as a passive handwriting pad to capture handwritten English letters and gestures with an unmodified commercial smartwatch. The core design rationale is that Swift measures high-frequency finger-on-skin friction acoustics and complements them with active ultrasonic probing to better recognize characters with similar stroke trajectories. A lightweight cross-attention fusion model is designed to reliably integrate the two modalities, while robustness is ensured via various augmentation techniques. Across an offline technical evaluation in lab and everyday environments (N=22), Swift achieves a macro F1-score of 0.91 after leave-one-subject-out training and few-shot personalization. A real-time deployment study (N=8) demonstrates phrase composition on the watch with users reporting the system as learnable and engaging. A Wizard-of-Oz comparison study (N=8) further indicates users generally prefer Swift over touchscreen handwriting, with preference increasing for more demanding scenarios such as heads-up text entry.
Authors
- Wentao Xie (ORCID: https://orcid.org/0000-0002-7858-4419)
- Yixuan Liu (ORCID: https://orcid.org/0000-0001-6529-3527)
- Jin Zhang (ORCID: https://orcid.org/0000-0002-2674-0918)
- Chi Xu (ORCID: https://orcid.org/0009-0000-7036-7148)
- Yankai Zhao (ORCID: https://orcid.org/0009-0003-0408-9137)
- Qian Zhang (ORCID: https://orcid.org/0000-0001-9205-1881)
- Z Qin
Institutions
- Hong Kong University of Science and Technology (HK)
- Southern University of Science and Technology (CN)
- University of Hong Kong (HK)
Publication Details
- Journal
- Proceedings of the ACM on Interactive Mobile Wearable and Ubiquitous Technologies
- Published
- 2026-09-30
- DOI
- https://doi.org/10.1145/3832006
- Primary Topic
- Interactive and Immersive Displays
- Type
- article
- Field-Weighted Citation Impact
- 0.00