EITWatch: ESP32-S3 Smartwatch for Planar EIT Gesture Recognition

Researchers at Northwestern University have developed EITWatch, an open-source Electrical Impedance Tomography gesture-sensing system built into a standard 40 mm smartwatch case. According to the research team, the device relies on a Seeed Studio XIAO ESP32-S3 board and an array of eight electrodes on the back of the watch case to detect hand gestures by measuring impedance changes caused by muscle and tendon movements beneath the skin.

Hardware Specifications and Analog Front-End Architecture

The EITWatch hardware centers on an Espressif ESP32-S3R8 system-on-chip, featuring a dual-core Tensilica LX7 microcontroller operating up to 240 MHz with vector instructions for AI acceleration. According to technical documentation from the developers, the board includes 512KB of SRAM, 8MB of PSRAM, and 8MB of SPI flash storage. Wireless connectivity is handled via dual-mode Wi-Fi 4 and Bluetooth 5.0 using an external u.FL antenna, while a USB Type-C port provides power supply and programming capabilities.

The sensing surface utilizes eight gold-plated stainless steel disk electrodes, each measuring 2 mm in diameter. These are arranged in a 31 mm diameter ring with approximately 11.9 mm of adjacent spacing, mounted on a 6-layer PCB measuring 60 x 40 mm that fits beneath a 40 mm watch housing. A 20 mm PCB extension holds the discrete analog front-end (AFE), which features an AD5930 waveform generator outputting a 50 kHz sinusoid, an ADA4841 voltage-to-current drive providing roughly 1 mA peak-to-peak, an AD8220 instrumentation amplifier, and an AD7450 12-bit SPI ADC. Drive current and sense gain are configured using AD5270 digital potentiometers, while four ADG738 8-channel matrix switches route the injection and sensing pairs.

Multi-Depth Scanning Protocols and Gesture Recognition Accuracy

Because a flat array cannot wrap around the wrist like traditional circumferential electrode bands, EITWatch employs a multi-depth scanning protocol. According to the research findings, the system keeps one electrode as a fixed source and cycles the negative sink through the remaining seven positions. This method allows the device to measure various depths and angular current paths, capturing 35 impedance measurements per frame at a rate of 48 Hz.

The ESP32-S3 runs an ESP-IDF (v5.2+) application that streams EIT data over Wi-Fi to a PC for logging and classification. The study reports gesture recognition accuracy reaches 92.5% for held macro-gestures and 91.5% for micro-gestures such as pinches and swipes. However, testing demonstrates that accuracy declines over time and across users. When tested 48 hours later, accuracy drops to 73.2% for macro-gestures and 70.4% for micro-gestures. When introduced to new users, accuracy falls further to 63.1% for macro-gestures and 55.3% for micro-gestures.

The research team tested six macro-gestures—Six, Gun, Point, Thumb Up, Stretch, and Fist—alongside five micro-gestures consisting of Swipe Left, Swipe Right, Pinch, Splay, and Wrist Flip, plus a relax class. Compared against prior open-source and DIY smartwatch builds like the MutantW V1, Sensor Watch Pro, and Core 2 Duo, EITWatch introduces integrated gesture recognition capabilities previously unsupported in those form factors.

Open-Source Availability and Resource Access

The EITWatch project is fully open-source. According to the development team, all hardware files, including Altium PCB design files, Gerbers, and the Bill of Materials (BoM), are released under the CERN-OHL-P-2.0 license. The C-based ESP32-S3 firmware is made available under the Apache-2.0 license. Developers and researchers can access all project files on the official EITWatch GitHub repository, review technical breakdowns on Xuanyou Liu’s website, or view a live demonstration posted on LinkedIn.

Frequently Asked Questions

What microcontroller powers the EITWatch?

EITWatch is powered by a Seeed Studio XIAO ESP32-S3 board featuring the Espressif ESP32-S3R8 system-on-chip.

How many electrodes does the EITWatch use for sensing?

The device uses eight 2 mm gold-plated stainless steel disk electrodes arranged in a 31 mm ring on the back of the watch case.

What is the gesture recognition accuracy of the system?

According to the research paper, initial testing shows 92.5% accuracy for macro-gestures and 91.5% for micro-gestures, though accuracy drops with new users and over time.

Where can I find the design files for EITWatch?

All hardware design files, firmware, and documentation are available on the project’s GitHub repository under open-source licenses.