Raspberry Pi 4B
Adrian Kingsley-Hughes/ZDNET
Follow ZDNET: Add us as a preferred source on Google.
ZDNET’s key takeaways
- Heartbeat data can be extracted from Wi-Fi CSI signals.
- Tested with ESP-32 modules and Raspberry Pi hardware.
- Accuracy reached 99.81% compared with medical methods.
Heart rate monitoring has been a mainstay of health and fitness, and while the stethoscope has given way to more precise electronic measuring devices, they’ve always involved wires and straps<!–>, or, in recent years, devices worn on the wrist–> or finger<!–>.
But what if there were a way to measure your heart rate wirelessly … using Wi-Fi?
Also: Internet running slow? 3 things I always check first for faster Wi-Fi speeds at home
This is exactly what researchers at the University of California have pulled off–> recently.
Non-intrusive, continuous monitoring system
Using the way Wi-Fi signals permeate the environment, they analyzed how a beating human heart affected the Channel State Information (CSI — part of the Wi-Fi signal that contains information about how the signal moves from the transmitter to the receiver).
They came up with a non-intrusive, continuous heart rate monitoring system called Pulse-Fi.
Also: I’ve tested the Apple Watch, Oura Ring, and other sleep trackers – 5 tips to get the best results
This was so interesting that I had to buy the paper to take a look, and it’s rather fascinating.
To verify their results, the group collected two datasets — one using two ESP-32 Wi-Fi modules<!–>, which retail for about $5, and another using the Raspberry Pi 4B–>, which costs about $30 to $50.
Also: The best Raspberry Pi alternatives of 2025: Expert recommended
The data collected using Wi-Fi was compared to data collected simultaneously using traditional methods.
After collecting the data — isolating the part of the CSI signal that relates to movements caused by a beating heart, removing environmental noise, applying a bandpass filter to target the 0.8 to 2.17 Hz range (corresponding to 48 to 130 beats per minute), and adding another filter to reduce noise while preserving signals needed — the data is crunched using a a low-compute Long Short Term Memory (LSTM)<!–> neural network, and a heartbeat is literally pulled out of the air.
But how accurate is Pulse-Fi?
Very accurate. The researchers found that using the ESP-32 modules, they could get an accuracy of 99.38%, or an error of 0.51 BPM compared to traditional heart rate measuring tech, and 99.81% when using the Raspberry Pi, an error of only 0.2 BPM.
That’s clinical-level accuracy.
The increased accuracy of the Raspberry Pi is due to the Wi-Fi module having more subcarrier frequency channels than the ESP-32 module (234 compared to 64). And this accuracy requires only five seconds of data collection, using only off-the-shelf parts.
Is Pulse-Fi going to replace how heart data is collected?
Right now, probably not, but this opens up new possibilities in both medical and health fields.
–>
