GaitGuard

GaitGuard is a wearable rehab system that gives patients real-time, personalized feedback on movement between physical therapy visits. A therapist first demonstrates the correct motion, the patient performs a few calibration reps, and the system builds a personalized digital twin of their correct movement. During future reps, GaitGuard compares live motion to that model and fires a haptic vibration if the movement is off - no screen-watching required.

The system consists of three custom wearable sensor units. The main unit straps to the shin and houses an ESP32, MPU-6050 IMU, LiPo battery with charger, and a haptic motor - all inside a custom 3D-printed SolidWorks enclosure printed in PLA. A second IMU unit straps to the ankle, and a third ESP32 unit with a 1.69" touch display attaches to the foot. Each ESP32 samples at 50 Hz and streams raw IMU data over WiFi to a Raspberry Pi 5, where a complementary filter fuses the signals into stable joint angle estimates corrected for gyroscope drift.

The core of the system is a two-layer stacked LSTM neural network built in PyTorch. Given the first portion of a new rep, the model predicts what the correct remainder should look like - trained entirely on the patient's own calibration data, not a population average. Deviation from the predicted trajectory is condensed into a Gait Health Score (0-100), and when it drops below a threshold, the haptic motor fires.

Critically, GaitGuard is not limited to walking. It can be configured for many different rehab exercises, making it a general-purpose tool for outpatient physical therapy.

Placed 3rd overall out of 650+ participants from several top universities at the Yale Hackathon (Design Sprint). Built entirely from scratch with a team of 4 in under 24 hours - no sleep.