ViBand: High-Fidelity Bio-Acoustic Sensing Using Commodity Smartwatch Accelerometers
summary
Custom kernel unlocks a 4,000 Hz accelerometer on commodity smartwatches and fuses axes (max across X/Y/Z), runs FFT and spectral/statistical features into an SVM. Built demos classify 17 gestures and recognize 27 grasped objects with about 94% accuracy across 17 users, see ~5% accuracy drop after watch removal, and only 6 false positives in 77 minutes. Structured-vibration transmission hits 1% bit error rate at 200 bps. Battery cost is nontrivial.
AI-written summary · about
chapters (8)
- 0:00Introduction to ViBand
- 1:32Overcoming sensing limitations
- 2:51Live demonstration
- 6:24Technical implementation
- 8:43Application domains
- 11:03Structured vibrations
- 13:23Evaluation and results
- 19:19Conclusion and Q&A
related talks
13mAcustico: Surface Tap Detection and Localization using Wrist-based Acoustic TDOA Sensing
06mDesigning and Evaluating Hand-to-Hand Gestures with Dual Commodity Wrist-Worn Devices
09mVHP: Vibrotactile Haptics Platform for On-body Applications
09mRotoWrist: Continuous Infrared Wrist Angle Tracking using a Wristband
20mSkinTrack: Using the Body as an Electrical Waveguide for Continuous Finger Tracking on the Skin
06mIdentifying Contact Fingers on Touch Sensitive Surfaces by Ring-Based Vibratory Communication
11mWrist2Finger: Sensing Fingertip Force for Force-Aware Hand Interaction with a Ring-Watch Wearable
15mBangle.js - an Open Source Smartwatch