Inspiration
Every 11 seconds, an older adult is treated for a fall injury. Most deaths aren't from the fall itself — they're from lying undiscovered for hours. We wanted to build something that works passively, requires no behavior change from the elderly, and costs under $100 to deploy anywhere.
What it does
GuardWave uses ESP32 nodes and WiFi DensePose to detect falls in real time through signal distortion — no cameras, no wearables. When a fall is detected, caregivers receive an instant alert. The system also learns normal movement patterns to flag early signs of deterioration.
How we built it
We built on (Pi)ruvnet's RuView — running the WiFi DensePose pipeline via Docker, with ESP32-S3 nodes capturing CSI data and streaming it via UDP to a Rust aggregator. Motion sensor data fuses with CSI features to reduce false positives and improve fall classification confidence.
Challenges we ran into
CSI data is incredibly noisy. Distinguishing a fall from someone sitting down fast took significant tuning. We also had no ESP32 hardware initially, so we validated the full pipeline using RuView's simulation mode and MM-Fi dataset replay before touching physical hardware.
Accomplishments that we're proud of
Getting fall detection working without a single camera or wearable felt like a real breakthrough. The entire system runs locally — no cloud, no subscription, no privacy tradeoff. And the hardware cost landed under $60, making it genuinely accessible for families without institutional budgets.
What we learned
WiFi signals are remarkably expressive — breathing, gait, and fall signatures are all readable once you understand CSI. We also learned that fusing a simple motion sensor with WiFi DensePose dramatically cuts false alarms, which is critical when alert fatigue is a real caregiver problem.
What's next for GuardWave
Predictive fall risk scoring based on gait pattern trends over days and weeks. Integration with SMS and smart home platforms for richer alerting. And a one-command installer so any family member — not just developers — can deploy GuardWave in a loved one's home in under 30 minutes.
Built With
- microcontrollers
- non-invasive-detection
- python
- sensor-fusion
Log in or sign up for Devpost to join the conversation.