PerCom 2026

μTouch: Enabling Accurate, Lightweight Self-Touch Sensing with Passive Magnets

Siyuan Wang1, Ke Li2, Jingyuan Huang1, Jike Wang1, Cheng Zhang2, Alanson Sample3, Dongyao Chen1

1Shanghai Jiao Tong University   2Cornell University   3University of Michigan

muTouch teaser overview
muTouch combines a compact 3-sensor magnetic unit with passive magnetic attachments to detect fine-grained face-touching and body-scratch behaviors.

Abstract

Self-touch gestures (e.g., nuanced facial touches and subtle finger scratches) provide rich insights into human behaviors, from hygiene practices to health monitoring. However, existing approaches fall short in detecting such micro gestures due to their diverse movement patterns.

We present muTouch, a magnetic sensing platform that combines a compact low-power hardware design, a lightweight semi-supervised pipeline requiring minimal user data, and an ambient field detection module for interference mitigation. In user studies, muTouch achieved 93.41% accuracy on eight-class face-touching detection and 94.63% on body-scratch detection, while requiring less than one minute of user fine-tuning.

Key Highlights

2.4 x 1.2 cm

Compact Hardware

Three low-power magnetometers in a miniaturized wearable sensing unit.

< 1 minute

Fast Personalization

Three samples per gesture are sufficient for rapid user adaptation.

93.41% / 94.63%

High Accuracy

Strong performance for both face-touching and body-scratch detection.

System Design

Complete muTouch hardware overview across mounting configurations and passive magnetic form factors

Complete Hardware Overview

Full system hardware demonstration, including the sensing unit, holder, glasses/earbud mounting, and passive magnetic front-end form factors.

muTouch hardware design

Wearable Sensing Hardware

Simulation-guided hardware design selects a three-sensor layout, balancing sensing quality with small form factor for glasses, earbuds, and skin-mounted usage.

MagDelta interference mitigation

MagDelta Interference Mitigation

A training-free trigger estimates background field and detects nearby magnets, improving robustness under environmental magnetic bias.

Applications and Evaluation

11

Participants (Face-Touching)

12

Participants (Body-Scratch)

90.50%

Follow-up Accuracy (Face)

92.59%

Follow-up Accuracy (Scratch)

BibTeX

@misc{wang2026mutouchenablingaccuratelightweight,
  title={{\mu}Touch: Enabling Accurate, Lightweight Self-Touch Sensing with Passive Magnets},
  author={Siyuan Wang and Ke Li and Jingyuan Huang and Jike Wang and Cheng Zhang and Alanson Sample and Dongyao Chen},
  year={2026},
  eprint={2601.22864},
  archivePrefix={arXiv},
  primaryClass={cs.HC},
  url={https://arxiv.org/abs/2601.22864}
}