Privacy-first, edge-AI elderly monitoring system that detects falls in real time.
Built for the Arduino UNO Q — all processing happens on-device, no cloud required.
- Real-time fall detection — MediaPipe pose estimation running locally
- Privacy first — Video never leaves the device
- Camera tracking — Pan/tilt servos follow the person
- Smart escalation — Layered verification before contacting emergency services
- Live monitoring — Family can view feed via local PWA
- Push notifications — Alerts via ntfy.sh (no app needed)
┌─────────────────────────────────────────────────────────────────┐
│ ARDUINO UNO Q │
│ ┌──────────────────────────────┐ ┌─────────────────────────┐ │
│ │ Qualcomm QRB2210 (Linux) │ │ STM32U585 (MCU) │ │
│ │ │ │ │ │
│ │ • MediaPipe Pose │ │ • Servo control │ │
│ │ • Fall detection │◄─► • Buzzer alerts │ │
│ │ • Web server │ │ • Button input │ │
│ │ • Notifications │ │ • LED status │ │
│ └──────────────────────────────┘ └─────────────────────────┘ │
└─────────────────────────────────────────────────────────────────┘
│ │
▼ ▼
USB Camera Physical Controls
(Logitech Brio) (Button, Buzzer, Servos)
Person Falls
│
▼
┌─────────────────────┐
│ COUNTDOWN #1 │ ◄─── Buzzer beeping
│ (30 seconds) │ LED flashing
└─────────┬───────────┘
│
┌─────────┴─────────┐
│ │
[BUTTON] (timeout)
│ │
▼ ▼
DISMISS ✓ Family Notified
│
┌─────────┴─────────┐
│ │
┌─────────▼─────────┐ │
│ COUNTDOWN #2 │ [APP DISMISS]
│ (60 seconds) │ │
│ • +30s if │ │
│ person stands │ ▼
└─────────┬─────────┘ DISMISS ✓
│
(timeout)
│
▼
Emergency Services
| Principle | Implementation |
|---|---|
| Local Storage | User information stored locally on-device |
| Infrared Camera | Additional anonymity — no identifiable video |
| Confidence-Based Escalation | AI confidence scoring before triggering alerts |
| Multi-Layered Verification | Human verification at each step before contacting emergency services |
| No Cloud Processing | All AI runs on-device via edge computing |
| Local Network Streaming | Live feed only accessible on home WiFi |
| Component | Purpose |
|---|---|
| Arduino UNO Q (4GB) | Main processing unit |
| USB Webcam | Video input |
| 2x Servo Motors | Pan/tilt camera tracking |
| Push Button | "I'm OK" dismiss |
| Piezo Buzzer | Audio alerts |
Family members receive push notifications and can view a live feed through a Progressive Web App (PWA) on their phone — no app store download required.
Features:
- Real-time video stream
- "I'm OK" remote dismiss
- One-tap emergency call
- Connection status indicator
- Download repo as .zip file
- Connect to Arduino Uno Q with network mode
- Attach USB dongle and USB webcam
- Import .zip as App Lab project into Arduino App Lab
- Camera fall detection with AI
- Fall detection signal (buzzer + LED)
- Signal off button
- Inactivity emergency contact escalation
- Camera movement tracking (pan/tilt)
- Print camera housing
- Circle member viewing UI with video playback
| Name | Role | GitHub |
|---|---|---|
| Adam Le | Software | @adamvl7 |
| Kevin Chhim | Embedded | @kevinlycc |
| Ryan Ong | Project Manager | @riannongg |
| Samuel Phan | Machine Learning & Systems | @blayyd |
- https://docs.edgeimpulse.com/hardware/deployments/run-arduino-app-lab
- https://docs.arduino.cc/hardware/uno-q/
- https://github.com/arduino/app-bricks-py
Built with love ❤️ at UCI IrvineHacks 2026
