Northwestern Researchers Unveil 92.5% Accurate ESP32-S3 EITWatch in 40 mm Smartwatch Case
Updated
Updated · CNX Software · Aug 14
Northwestern Researchers Unveil 92.5% Accurate ESP32-S3 EITWatch in 40 mm Smartwatch Case
1 articles · Updated · CNX Software · Aug 14
Summary
EITWatch packs planar electrical impedance tomography into a standard 40 mm smartwatch form factor, using eight rear-case electrodes to recognize hand gestures with 92.5% macro-gesture accuracy and 91.5% for micro-gestures.
The design replaces wraparound wrist electrodes with a 31 mm flat electrode ring and a multi-depth scanning method that captures 35 impedance measurements per frame at 48 Hz.
An ESP32-S3 handles both streaming and fully on-device inference; the standalone pipeline draws about 35 mA at 4.3 V, yielding roughly 8.6 hours from a 300 mAh battery.
Performance weakens over time and across users, dropping after 48 hours to 73.2% for macro-gestures and 70.4% for micro-gestures, and to 63.1% and 55.3% for new users.
Northwestern released the hardware and firmware as open source, including PCB files, Gerbers and BoM, making EITWatch a rare DIY smartwatch platform with built-in gesture recognition.