Scientists invent ‘Pulse-Fi’ prototype — a Wi-Fi heart rate monitor that’s cheaper to set up than the best wearable devices
The engineers used artificial intelligence (AI) and inexpensive, commercially available hardware to convert the amplitude of Wi-Fi signals into estimates of a person’s heart rate.
The accuracy of this system, called Pulse-Fi, is remarkably consistent across body positions and distances, the researchers wrote in a study published Aug. 5 in the Proceedings of the 2025 IEEE International Conference on Distributed Computing in Intelligent Systems and the Internet of Things (DCOSS-IoT).
Many home technologies, such as chest strap monitors and smart watchesmonitor vital signs, including heart rate and respiratory rate. However, these devices require constant contact with the individual and are expensive, necessitating contactless technologies.
One of these technologies can exploit the information contained in Wi-Fi signals, which are radio waves which transport data between a transmitter and a receiver, for example between a router and a computer.
“Channel State Information” (CSI) provides the amplitude and phase of the signal as it passes between these two devices, including when passing through obstacles such as moving chests. Because signals distort as they pass these barriers, researchers can filter CSI data to capture vital signs.
Miscellaneous examples now exist for Wi-Fi heart rate detectionBut Kocheta and his team argued that several limitations remain. For example, many rely on hardware that is now defunct. To address these limitations, researchers developed a new system called “Pulse-Fi.”
Capture vital signs
To collect the data needed to evaluate Pulse-Fi, the team placed seven people – five men and two women – between two single antennas. ESP32 devices. These microcontroller devices emitted Wi-Fi signals, with one acting as a transmitter and the other as a receiver. Participants’ actual heart rates were recorded at the same time via a pulse oximeter attached to their fingertip.
Each individual participated three times: once at 3.3 feet (1 meter) from the EPS32s, then at 6.6 feet (2 m) and 9.8 feet (3 m). Each measurement window lasted five minutes.
The team then developed a machine learning pipeline to estimate CSI heart rates. The first step was to extract amplitude information related to individual heartbeats, then remove disordered parts of the signal originating from obstacles in the environment.
Next, engineers added a filter to remove signal frequencies outside the range of 0.8 to 2.17 hertz, which corresponded to 48 to 130 beats per minute (BPM). Then they added a second filter to further smooth the signal.
The team then estimated the participants’ heart rates using a long-term and short-term memory recurrent neural networka form of machine learning that adds “memory cells” to the processing of sequential data, which provides the context needed to detect dependencies in the data. In this case, these addictions are linked to things like resting heart rate and exercise-induced BPM spikes.
The team was surprised to find that heart rate estimates remained accurate across different distances from the ESP32 devices. Pulse-Fi underestimated and overestimated heart rate by an average of 0.429 BPM at 1 meter, 0.482 BPM at 2 m, and 0.488 BPM at 3 m.
The researchers then used pre-existing data CSI Wi-Fi Health Data to test the performance of Pulse-Fi with different body positions and activities. The data comes from 118 Brazilian adults occupying 17 stationary and active positions, including sitting, walking in place, and sweeping the floor for 60 seconds. Participants were 3.3 feet (1 m) from the Wi-Fi transmitter and receiver as well as the Raspberry Pi 3B+ used to collect CSI data.
They compared the neural network’s heart rate estimate to the smartwatch readings and found that Pulse-Fi was not affected by the person’s body position. The typical error was 0.2 BPM.
Wireless beats
This early technique is theoretically interesting, said Andreas Karwatha health data scientist at the University of Birmingham in the United Kingdom who was not involved in the research.
However, he said a major limitation of this research is that the same data was used for training, validation and testing of the model. The researchers mixed up the data each time, but Karwath said that created a self-fulfilling prophecy.
“It’s like predicting a person’s illness by learning from the person and then predicting the person,” he told Live Science. “It doesn’t make sense.”
In response to this criticism, the researchers said that although their analysis involved shuffling, they have since tested the model in real time, where the Pulse-Fi was trained only on past data and then evaluated on a completely new input signal and environment. This research has not yet been published.
Karwath also explained that the smartwatch and oximeter used to collect heart rate information for the neural network to compare are not always 100% accurate, so their data can be biased.
Kocheta, Bhatia and Obraczka recognized this limitation of the smart watch. However, “the pulse oximeter is generally considered a certified and highly accurate medical device,” they said.
The team is currently expanding Pulse-Fi testing to track the heart rates of multiple people in a room at the same time to see how well the model copes with crowded environments.
The authors stated that no explicit personal information is involved in the data processing pipeline and that all heart rate estimates remain in the hardware. As such, this technology does not pose any data privacy concerns. Karwarth predicted the technology will be deployed in at least five to 10 years.
