'ESPectre' turns your home Wi-Fi into a motion sensor without machine learning and integrates it with Home Assistant
How I Turned My WI-FI Into A Motion Sensor | Francesco Pace | Medium
https://medium.com/@francesco.pace/how-i-turned-my-wi-fi-into-a-motion-sensor-61a631a9b4ec
GitHub - francescopace/espectre: ESPectre 👻 - Motion detection system based on Wi-Fi spectrum analysis (CSI), with Home Assistant integration.
https://github.com/francescopace/espectre
Mr. Pace, a home automation enthusiast, didn't wait for devices compatible with IEEE 802.11bf, which standardized Wi-Fi sensing, to appear, but instead developed his own motion detection system, 'ESPectre,' based on spectrum analysis of Wi-Fi signals. The features of ESPectre are as follows:
- It uses the ESP32-S3, a microcontroller that costs less than $10 (about 1,500 yen).
- It can be integrated with Home Assistant , which allows you to manage IoT devices together and create a smart home.
-Can detect movement even through walls.
- It uses only mathematics and signal processing, not AI or machine learning.
'The problem with Wi-Fi sensing is that there's a lot of noise floating around in the air, and most existing solutions use machine learning to clean up the data and identify patterns of human movement. But I wanted to demonstrate that by using only mathematics and signal processing, we can achieve good results even with inexpensive hardware,' he said.
Pacce explains Wi-Fi sensing as 'like holding your hand in front of a flashlight in a dark room and watching the shadow it casts.' Instead of light, Wi-Fi sensing uses the Wi-Fi signal that travels between the router and the sensor. When this signal is disrupted by the movement of a person or object, the sensor can detect this movement.
In modern Wi-Fi systems, dozens of different frequencies (subcarriers) are transmitted in parallel, and the receiver calculates the signal strength (amplitude) and phase (phase) of each frequency. When a person moves within the signal range,
By using mathematical component analysis, digital filters, feature extraction, etc., Pace succeeded in developing 'ESPectre' without using machine learning. ESPectre can detect someone's presence with about 95% accuracy, and Pace reports that the latency is less than 50 milliseconds per packet.
The ESPectre device part created by Pacce looks like this: It is a very compact device with an ESP32-S3 microcontroller connected to an external antenna to improve reception sensitivity.
Possible uses for ESPectre include as a home security device that notifies you when an object is detected, an automation function that turns on lights only when people are present, an energy-saving function that turns off the air conditioner in an empty room, and a monitoring function that notifies you when an elderly person has not moved for an extended period of time.
However, ESpectre's limitations include its lack of machine learning algorithms, which means it struggles to distinguish between subtle movements and objects. For example, it can't distinguish between humans and pets, get an accurate count of the number of people in a room, or distinguish between walking and running.
Furthermore, while the Channel State Information (CSI) data used for Wi-Fi sensing is technically anonymous, it can detect people and movement, which poses risks of unauthorized surveillance and privacy violations. For this reason, Pacce recommended using ESpectre only in environments you can manage yourself, such as at home, and not installing it in shared spaces, to respect the privacy of visitors.
Pacce himself wrote a thread on the social news site Hacker News reporting on ESpectre.
Show HN: ESPectre – Motion detection based on Wi-Fi spectrum analysis | Hacker News
Pacce says he plans to use ESpectre to create a Wi-Fi Theremin, a theremin -like instrument played by holding your hand near the antenna. The key is turbulence fluctuations, which are continuous and stable, making them ideal for mapping to pitch and frequency. Pacce has two children, so he thinks the Wi-Fi Theremin could be used for games like hide-and-seek.
Related Posts:
