Overview
Parkinson's disease causes tremors and slowness of movement that make everyday tasks difficult, yet most patients are evaluated only a few times a year, and existing tremor-relief devices are expensive. We set out to build an affordable wearable that does three things at once: detect and quantify tremors, actively stabilize the hand when they happen, and track other Parkinson's motor symptoms between doctor visits.
Detecting tremors with machine learning
- Trained a machine learning classifier on the Parkinson's Disease Smartwatch dataset to distinguish tremors from voluntary hand movement
- Split live motion data into short overlapping windows and extracted statistical time-domain features (RMS, standard deviation, peak-to-peak amplitude, zero-crossing rate) from the accelerometer and gyroscope
- Deployed the trained model onto a Raspberry Pi Zero 2 W so classification and severity scoring run in real time on the device itself
Counteracting tremors in real time
- Integrated servo motors with the classification system so the model's predictions trigger a physical response
- When a tremor is detected, the servos pull elastic bands across the back and palm of the hand, adjusting tension to mechanically stabilize it, then release when the tremor subsides
- Designed the glove and electronics housing in Fusion 360 and 3D printed them in flexible TPU and rigid PLA
Tracking symptoms beyond tremors
- Built a companion mobile app that receives live tremor classifications and severity scores from the glove
- Designed bradykinesia assessments (timed finger-tapping tests) that measure motor slowness using statistical features of tap speed and rhythm
- Visualizes tremor activity and symptom trends over time so patients and clinicians can follow progression between appointments
Testing
We evaluated the prototype on simulated up-and-down, side-to-side and rotational ("pill-rolling") tremors modeled on published tremor characteristics, comparing motion with the stabilizing mechanism on and off and validating the results with paired statistical tests.
Publication and recognition
Our paper has been submitted to The Bulletin of the New Jersey Academy of Science and is under scientific review.
- 2nd Place, Engineering, and the U.S. Air Force Award for Excellence in Science & Engineering, BCA Research Expo (2026)
- 3rd Place, Engineering, and the Intelligent Systems for Real-World Impact Award, New Jersey Science Research Symposium (2026)
- 3rd Place, Engineering, New Jersey Academy of Science Junior Academy (2026)
- Qualified for the American Junior Academy of Science