InviSense

Home Page

Inspiration

Many elderly live alone, and the usual ways to check on them either feel intrusive (cameras), unreliable (panic buttons), or annoying to maintain (wearables). We wanted something that gives families peace of mind without watching or recording anyone. Also, it’s Hack&Roll, so we added a fun twist: we turned the lanyard into a sensor too.

What it does

InviSense turns home Wi Fi into a motion sensor. ESP32 devices “listen” to Wi Fi CSI changes and convert them into a movement score. The dashboard shows:

Live activity spikes (movement vs idle)

Per sensor thresholds you can tune

A floor plan view with draggable sensor placement and live metrics The goal is to support inactivity detection for elderly living alone, and alert family if activity stops for an unusual period.

How we built it

ESP32 firmware (based on ESPectre CSI motion detection algorithms) to collect CSI and output movement metrics

Raspberry Pi backend (Python + FastAPI) to aggregate and stream sensor data

MQTT + WebSockets for low latency real time updates

React dashboard (TypeScript + Vite + Tailwind) for the live charts, threshold controls, and floor plan mapping

Challenges we ran into

Signal noise and environment differences: Wi Fi behaves differently depending on room layout, distance, and interference.

Getting stable “baseline” readings: movement detection is only good if the initial calibration is clean.

Real time streaming: keeping the dashboard responsive with multiple sensors updating frequently.

Making it understandable: translating “weird signal data” into UI that a caregiver can actually use.

Accomplishments that we're proud of

Built an end to end system: ESP32 → Pi backend → real time dashboard

Achieved sub second live updates of movement score and sensor status

Designed a UI that’s actually usable: threshold settings + overview + floor plan

Kept it privacy first by design: no cameras, no audio, no wearables

Bonus Hack&Roll energy: lanyard as a sensor

What we learned

CSI based sensing is powerful, but calibration and placement matter a lot

Real time systems are mostly about handling edge cases and stability

What's next for InviSense

Ship the notification system end to end so caregivers reliably get alerted (Telegram/WhatsApp/email) when inactivity is detected.

Test in real homes across different room sizes, layouts, and router placements to validate reliability and reduce false alarms.

Make alerts meaningful across different sensor sensitivities by adding auto scaling / normalisation: if a sensor’s movement score naturally peaks at 30%, we rescale that 0–30% range to 0–100% so the alert threshold slider stays consistent and intuitive.

Improve robustness with better calibration routines and multi sensor coverage for fewer dead zones.