WiFi Sense extracts raw 802.11 CSI to detect human motion and respiration. No cameras. No radar. Just radio waves.
Hover your mouse over the radar field to distort the WiFi subcarriers in real-time.
Isolates tiny Doppler shifts. A chest moving just 6mm during breathing is perfectly resolved at 0.13–0.7 Hz.
No optical sensors. Operates in total darkness. Detects presence without capturing images.
CSI data is routed over an AES-128 encrypted nRF24 radio link to prevent measurement corruption.
100Hz UDP stream pushed to the ESP32. Raspberry Pi handles filtering, peak picking, and a live WebGL dashboard.
> git clone https://github.com/The-Masked-Bear/wifisense-pi
> cd wifisense-pi && python3 -m venv .venv
> ./.venv/bin/pip install -r requirements.txt
> cd pi && ../.venv/bin/python -m wifisense --link synthetic
> DASHBOARD LIVE ON PORT 8080
System loaded. Type 'help' for available commands.
No.
An ESP32 microcontroller (~$3), a Raspberry Pi 4 (~$35), and an nRF24L01+ radio module (~$1). Total cost under $40. The ESP32 extracts raw CSI from WiFi frames, the nRF24 relays it encrypted, and the Pi runs all the DSP and the live dashboard.
Yes. WiFi operates at 2.4 GHz, which penetrates drywall, wood, and most interior building materials. The CSI (Channel State Information) captures how the wireless signal is distorted by objects and movement in the environment, including through walls. Concrete and metal will attenuate the signal but detection is still possible at reduced range.
Absolutely not. This is not a camera. WiFi CSI detects motion and presence only — it can tell that someone is in a room and whether they are breathing, but it cannot identify who they are, what they look like, or what they are doing. It is fundamentally a motion and vibration sensor, not an imaging system.
Yes. WiFi Sense performs entirely passive monitoring of standard 802.11 WiFi signals. It does not inject packets, intercept communications, or decrypt any data. It only analyzes the physical-layer channel state of its own transmitted signals. This is equivalent to measuring signal strength, which all WiFi devices do internally.
The system resolves chest displacement down to approximately 6mm using sub-carrier covariance analysis across 52 OFDM sub-carriers. Normal adult breathing produces 4–12mm of chest movement, well within detection range. The DSP pipeline isolates the 0.13–0.7 Hz respiration band from other motion artifacts.
Yes. All CSI data transmitted over the nRF24L01+ radio link between the ESP32 and the Raspberry Pi is encrypted with AES-128. This prevents anyone from intercepting or tampering with the raw sensor data in transit.