Wi-Fi has a weird new smart-home use case: detecting people

Wi-Fi normally connects your smart home, but it can also sense whether anyone is inside it. The trick isn't tracking a phone or watching connection logs. Instead, there are systems that can observe how a human body changes radio waves moving through a room. This idea is no longer a research demo being run in a lab either. Researchers have used Wi-Fi for everything from activity recognition to human identification, and routers are already turning into motion sensors in the real world.

And now, it's becoming accessible to Home Assistant users through cheap ESP32 boards, ESPHome, and software like TOMMY. The result is similar to a mmWave presence sensor, but its operating model, coverage, and trade-offs are surprisingly different.

Your router has been quietly watching you breathe

Radio waves bounce off walls, furniture, and, inconveniently, you

A Wi-Fi signal rarely takes one clean, direct route between two devices. Along the way to its destination, it bounces off walls, furniture, and more importantly, people. That produces multiple paths that arrive at slightly different times and strengths. Walking into that environment changes those paths because the body reflects, absorbs, and scatters some of that radio energy. Even breathing or shifting in a chair can produce small but detectable changes.

The metric that maps this is called Channel State Information, or CSI. On an ESP32, CSI describes the channel's frequency response across the individual subcarriers used by an OFDM Wi-Fi transmission. It provides much more detailed information than a simple signal strength indicator, including amplitude and phase-related channel data. Espressif exposes APIs that let compatible ESP32 chips collect CSI whenever packets are received, which turns an otherwise simple Wi-Fi microcontroller into a basic radio-sensing node.

Then, software can compare those measurements over time. A stable CSI pattern suggests that the environment is relatively still, while a changing pattern can indicate movement. With sensitive hardware and suitable processing, recurring micro-movements such as breathing can support stationary presence detection. The system isn't identifying a connected device carried by someone; it's detecting the physical effect of a body on a radio channel.

CSI isn't mmWave, and pretending otherwise ends badly

Different radios, different frequencies, very different opinions of walls

Wi-Fi sensing and mmWave often solve the same smart-home problem, but they're not the same thing. A mmWave presence sensor is a purpose-built radar, usually operating at 24, 60, or 77 GHz. It transmits a signal and studies its reflections to estimate motion and, depending on the hardware you have, details such as range, velocity, and angle. Its sensitivity to tiny movements makes it useful for keeping the lights on while someone sits at a desk.

CSI sensing, on the other hand, uses ordinary Wi-Fi signals in the 2.4 GHz or 5 GHz bands. These lower-frequency signals generally travel farther and penetrate common obstacles more readily, so a pair of nodes can sense across furniture or walls. That eliminates line-of-sight placement, but it also makes boundaries less precise. A person walking in a hallway may disturb a link intended to cover the living room. Current ESP32 boards also have limited antenna and channel capabilities, so claims about exact localization, pose detection, or reliable people counting should be treated differently from straightforward motion and occupancy detection.

TOMMY turns your pile of spare ESP32s into a presence sensor

Two boards, one band, and a server that has to stay online

TOMMY packages CSI sensing into something resembling the average smart-home product. A zone needs at least two supported ESP32 devices on the same Wi-Fi band, positioned at least one meter apart. The devices exchange small packets, and the receivers extract CSI from those transmissions. TOMMY's server then analyzes the data collectively and exposes the zone to Home Assistant or a Matter-compatible platform as a sensor.

If you've already stacked your home with ESPHome hardware, this is especially useful. TOMMY supports the original ESP32, C3, C5, C6, S2, and S3 families. A board can run dedicated TOMMY firmware, but the external ESPHome component can add sensing without removing the device's existing job. This means all your existing ESP32-based smart devices can be transformed into presence detection nodes without adding any new hardware.

There are technical catches, though. The ESPHome build must use ESP-IDF because CSI support depends on it, and TOMMY changes the relevant SDK configuration options. Its external component is downloaded from the local TOMMY server during compilation, so that server must be available. On mesh Wi-Fi, nodes belonging to one zone must connect to the same access point, which may require pinning them to a specific BSSID.

Cheap chips detect motion, only newer ones notice you sitting still

Stationary presence needs a C6 or C5, and a lot of patience

Only the ESP32-C5 and C6 can detect stationary presence. The original ESP32, C3, S2, and S3 are limited to motion detection, which makes them better deployed as motion-sensing endpoints than direct replacements for a good mmWave sensor. TOMMY lists the ESP32-C6 on the 2.4 GHz band as capable of supporting two or more devices, and the newer, dual-band ESP32-C5 on 5 GHz with four or more devices as the optimal setup. For stationary presence, the developer recommends a mesh topology, where nearly every device transmits for fuller coverage, and a hold time of around 10 seconds.

Placement matters too because a TOMMY zone covers the space between its nodes, regardless of room boundaries. Sure, more links will improve coverage, but they can also collect disturbances from pets, fans, curtains, or people behind a wall. TOMMY doesn't inherently know that a radio disturbance came from a human, so tuning thresholds and physically arranging nodes remain part of the installation.

Clever, logical, and definitely not a mmWave killer

Closed-source, trial pauses, and a server that phones home

The biggest appeal here is reuse. A few ESP32 boards, or devices already running ESPHome, can add occupancy-driven lighting and HVAC automations without cameras or a radar module in every room. Sensing data stays on your local network, and the server can run as a Home Assistant app or Linux Docker container. It is, however, closed-source at its core, although the Home Assistant integration is open source. The free trial pauses detection for one minute out of every two, and the paid edition uses a one-time license, while the server still requires internet access for license verification.

That makes Wi-Fi CSI an intriguing complement to PIR and mmWave rather than a universal replacement. mmWave is still a good choice when a room needs tightly controlled coverage or detailed ranging. Wi-Fi sensing makes more sense when broad coverage, hidden nodes, through-wall detection, low hardware cost, and ESPHome reuse matter more. Nothing visibly senses the room. The same radios that already carry your sensor readings and automation commands become the sensor. With realistic expectations and careful placement, Wi-Fi can tell a smart home not only that there are connected devices, but also that somebody is actually there.