No Camera. No Tag. Just Wi-Fi Sensing: Room Presence Detection with Microcontroller

Most presence detection systems need a camera, a PIR sensor, or something worn by the person. This talk asks a simpler question: what if the Wi-Fi signal already in the room is enough?

A single ESP8266 microcontroller costing under two hundred rupees can detect whether a person is in a room using nothing but the Wi-Fi signal around it. No camera. No wearable. No dedicated sensor. This works because a human body physically disturbs the radio environment it stands in.

Wi-Fi signals do not travel in straight lines. They bounce off walls, furniture, and the ceiling, arriving at the receiver from multiple directions at once. Engineers have called this multipath propagation and treated it as interference for decades. This talk takes the opposite approach. Instead of filtering those reflections out, it reads them. When a person walks into a room, their body absorbs and redirects Wi-Fi signals differently than empty air does. That change shows up as an increase in the short-term variance of RSSI readings on the ESP8266. When that variance, computed over a sliding window of samples, crosses a calibrated threshold, presence is detected.

This is not a standalone hobby project. In 2024, IEEE ratified 802.11bf, the Wi-Fi Sensing standard. It defines how ordinary Wi-Fi infrastructure can be used for presence detection, fall detection, breathing rate monitoring, and building occupancy sensing, all without cameras. The ESP8266 demo in this talk builds the same core idea from scratch using only the Arduino core, a WebSocket library, and a browser dashboard.

The session covers the full system. Firmware reading RSSI ten times per second. A variance algorithm running on the chip itself. WebSocket streaming to a live browser graph. And the calibration process that makes detection thresholds actually meaningful in a real room.

The parts that failed get equal time. RSSI noise, false triggers from a ceiling fan and a door opening, and baseline drift that made the system useless when moved to a different room. Those problems and the specific firmware decisions that addressed them are the most useful part of the talk.

It ends with a live demo. The phone moves closer. The graph changes. The audience sees it happen in real time.

All source code is on GitHub under the MIT License before the talk date.

1. Wi-Fi signals bounce off walls, furniture, and ceilings before reaching a receiver. A human body changes those bounce patterns in ways the hardware can actually measure. This talk explains that physics using a simple analogy. No RF background needed.

2. Device-free passive detection means the person being detected carries nothing. No phone, no tag, no beacon. This talk explains what that means technically, why it matters for privacy, and how it is different from phone-tracking or Bluetooth-based systems.

3. In 2024, IEEE ratified 802.11bf, a standard that formally turns Wi-Fi infrastructure into a sensing platform. This talk explains what that standard actually does, then shows a working prototype of the same idea built on a chip that costs under two hundred rupees.

4. You will leave with working open-source code for RSSI-based presence detection on ESP8266, covering the sliding window standard deviation algorithm, threshold calibration, and WebSocket streaming to a browser dashboard.

5. RSSI is noisy. CSI is richer but needs different hardware. That constraint was not just a footnote. It changed how every piece of the firmware was written. This talk walks through those decisions as a real example of what hardware-software co-design looks like in practice.

6. The first version of this system triggered false positives every time a ceiling fan turned on. This talk covers three failure modes that actually happened during development: RSSI noise, environmental movement triggers, and baseline drift. Each one and the firmware fix that resolved it.