A wristband that recognizes hand gestures by sensing how electricity flows through the wrist, then sorts them with machine learning. I built it end to end, from the circuit board to the model.
- When
- 2021–2023
- Context
- Independent project
- Technical summary
- A 4-layer printed circuit wristband using tomography to map electrical impedance for real-time gesture recognition via SVM.
- Topics
What it is
When you move your hand, the muscles and tendons in your wrist shift, and that changes how electricity passes through the tissue. This wristband measures those changes from many angles at once (a technique called electrical impedance tomography, or EIT) and uses them to tell which gesture you’re making, in real time.
What I did
This was an independent project from December 2021 to August 2023, and I handled every stage:
- Research: read up on EIT and wearable sensing
- Hardware design: schematic, PCB layout and component selection
- Fabrication: had the boards made and assembled them
- Software: firmware, signal processing and training the ML model
- Testing: user studies, then refining the algorithm
How it works
Hardware
- PCB: custom 4-layer printed circuit board
- Electrodes: an array of electrodes around the wrist, measuring impedance along many paths
- Microcontroller: low-power ARM processor for signal processing and wireless communication
- Power: rechargeable LiPo battery
Software
- Features: time- and frequency-domain features extracted from the impedance signals
- Classifier: a Support Vector Machine (SVM) trained on data from multiple users performing standard gestures
- Result: reliable classification of 6 distinct gestures
Status
Complete. Possible uses include hands-free device control, accessibility tools for people with limited mobility, and gaming controllers.