ESP32-S3 captures CSI
The board receives Wi-Fi traffic and extracts Channel State Information from incoming packets.
Wi-Fi CSI
SWMD
Detecting human presence through Wi-Fi, not cameras.
A privacy-first sensing system that reads tiny changes in Wi-Fi signals to infer room occupancy.
Built with ESP32-S3 CSI capture, a Python detection backend, and a live radar-style dashboard.
rate
~10 Hz
mode
CSI RX
status
occupied
About the project
Presence detection without cameras or wearables.
SWMD uses ordinary Wi-Fi signal behavior as a room sensor, keeping the sensing layer lightweight and privacy-conscious.
The system uses Wi-Fi Channel State Information (CSI) from an ESP32-S3 to detect when a person is present in a room, without any camera or wearable.
The ESP32-S3 captures CSI from incoming Wi-Fi packets. To keep a steady receive rate, it pings the gateway 10 times per second, then sends each CSI frame to a Python backend over the USB serial console.
The backend runs a presence detector that co-confirms motion, phase, and baseline shifts using auto-tuned thresholds, Hampel/PCA CSI denoising, and hold-and-decay logic.
The current senior-project flow uses one ESP32-S3 sensing node: calibrate the empty room, stream CSI to the laptop, and display a live occupied or empty decision on the dashboard.
Sensor
ESP32-S3 CSI
Privacy layer
No camera
Sampling target
10 frames/s
Output
Presence
How it works
A simple signal pipeline from room to dashboard.
CSI frames move from the ESP32-S3 to the PC, where the backend turns signal changes into an occupancy estimate.
USB serial to PC
Frames stream through the ESP32-S3 UART console while gateway pings keep the receive rhythm steady.
Python detects + visualizes
The backend filters CSI frames, applies the detector, and visualizes live confidence on a radar-style dashboard.
Tech stack
Hardware, firmware, backend, and dashboard.
The project combines embedded Wi-Fi CSI collection with Python signal processing and a browser-based visualization layer.
Hardware
ESP32-S3
Wi-Fi 2.4 GHz
CH343 UART bridge
Firmware
ESP-IDF v5.3
C
gateway ping loop
Backend
Python 3
Flask
NumPy
pyserial
Hampel/PCA filter
Frontend dashboard
HTML5 Canvas
vanilla JS
Tools
Git
VS Code
Windows PowerShell
Features
Built for real-time sensing and field testing.
SWMD focuses on live feedback, one-node calibration, and repeatable presence testing.
ESP32-S3
Gateway AP
One ESP32-S3 captures CSI from gateway replies and streams frames to the backend.
- Real-time presence detection (~10 Hz)
- Live radar-style room visualization
- Auto-calibration with user-controlled cooldown that lets you leave the room first
- Single ESP32-S3 sensing node
- CSI heatmaps for amplitude, baseline shift, and phase
- Hampel + PCA CSI denoising before detection
- USB serial CSI stream from the ESP32-S3 to the laptop
Prototype
Single-node CSI sensing, live backend feedback.
The prototype streams CSI from the ESP32-S3 into the laptop over USB serial, then Python detector code filters the frames and updates the radar dashboard.
ESP32-S3
CSI capture
USB serial CSI
Python processing
Python code
detector.py
ESP32-S3 node -> laptop USB serial stream -> Python detector code.
Hardware setup
One ESP32-S3 captures Wi-Fi CSI while placed inside the test room and connected to the laptop over USB.
Streaming link
The firmware prints CSI frames over the UART console, keeping the receiver rate stable with gateway pings.
Detection dashboard
The Python backend applies CSI denoising and detector logic, then displays activity on the live radar view.
Team
Students and supervisor.
S1
Sayed Jaafar Sadeq
Student
S2
Maitham Ali Isa
Student
S3
Hashem Saeed Alkhanaizi
Student
SV
Dr. Mohamed Baqer
Supervisor
Contact
Send us a message.
Contact us
Use the form below to reach the SWMD team directly.
SWMD was built as a compact sensing prototype for privacy-aware room presence detection using Wi-Fi CSI, embedded hardware, and Python-based inference.
info@wifisense.org
University of Bahrain