Presient Sensor — From contactless sensing to a working home automation

The question

A motion sensor can tell a home that someone is present. I wanted to explore a more ambitious question: could contactless physiological patterns help personalize a room without a camera or an explicit login?

That question led to Presient Sensor: a prototype spanning physical hardware, firmware, a backend, mobile enrollment and an automation layer.

The system I built toward

The sensor node sends presence and physiological estimates through MQTT. A backend compares readings with enrolled profiles and publishes an event. Home Assistant decides what device action follows.

The documented end-to-end demonstration controlled a media player. This establishes the integration chain; it does not establish reliable identification of a person. The enrollment interface, matching logic and automation controls let the larger idea be explored without presenting it as a finished authentication product.

Engineering decisions

Separate sensing, matching and action. The sensor produces observations; the backend evaluates them; the home automation layer applies policy. A hardware change should not require rebuilding every downstream integration.

Treat signals as observations over time. A single heart-rate reading cannot establish identity. The prototype explored simple statistical patterns and optional wearable input, while leaving actual identification accuracy an open research question.

Keep low-stakes automation reversible. Manual overrides and timeouts are part of the demonstration. Door access, computer unlocking and other authentication applications remain concepts requiring substantially different validation.

A failure that changed the design

An enrollment script reset the database each time it ran, erasing existing profiles. The reset became an explicit opt-in operation. That fix illustrates a practical boundary between a disposable development fixture and state that a working prototype must preserve. The source synthesis records the fix but does not cite an automated regression test.

What this demonstrates

Presient connects embedded hardware with an application system: sensor integration, messaging, enrollment, backend processing and real device behavior. It also shows why a successful integration demo and a validated biometric product are different milestones.

Role and limits

I conceived the product, selected and integrated the hardware and directed the software work. AI coding assistants helped implement it.

The documented radar module produced inconsistent heartbeat estimates and limited access to raw signals. A different radar platform was planned, not verified as completed. Multi-person interference and identification performance remain unresolved. This is neither a validated authentication system nor a medical device.