Pulse-Fi Tracks Heart Rates Through Ambient Wi-Fi Signals

Key Takeaways

• Pulse-Fi estimated heart rates with an error of less than 1.5 beats per minute in early testing.
• The system runs on low-cost ESP32 microcontrollers and Raspberry Pi devices without requiring wearables or cameras.
• University of California, Santa Cruz researchers are testing multiuser monitoring and considering commercialization.

Pulse-Fi, a contactless heart-rate-monitoring system developed at the University of California, Santa Cruz, uses ordinary Wi-Fi signals and an AI model to detect the minute physical changes associated with a heartbeat.

The approach could give healthcare providers, senior-living operators and wellness technology companies another way to monitor vital signs without asking people to wear a sensor. It also avoids the lighting requirements and some of the privacy concerns associated with camera-based monitoring.

A professor at the University of California, Santa Cruz led Pulse-Fi's development alongside a postdoctoral student and a high-school intern.

"Pulse-Fi uses ordinary Wi-Fi signals to monitor your heartbeat without touching you. It captures tiny changes in the Wi-Fi signal waves caused by heartbeats," the lead researcher said.

Wi-Fi transmissions already move through homes, hospitals, offices and assisted-living facilities. Pulse-Fi attempts to turn those existing radio waves into a sensing layer, rather than introducing another body-worn device.

The system filters background noise and isolates changes in Wi-Fi signal amplitude caused by a person's heartbeat. An AI model then interprets the filtered measurements and produces a real-time heart-rate estimate. Importantly for potential deployments, that processing can run on relatively modest edge hardware instead of depending on expensive clinical equipment or extensive cloud computing.

Results described at the 2025 International Conference on Distributed Computing in Smart Systems and the Internet of Things through IEEE Xplore covered two experiments. In the first, seven volunteers sat at distances of 1, 2 and 3 meters from two ESP32 microcontrollers running Pulse-Fi. Researchers compared the resulting estimates with measurements from a pulse oximeter.

The second experiment expanded testing to more than 100 participants and used Raspberry Pi devices. Participants walked, ran in place, sat down and stood up, allowing the research team to examine how movement and posture affected performance.

Pulse-Fi recorded an error of less than 1.5 beats per minute compared with the reference measurements. It also retained sufficient accuracy across different postures and at distances of up to 10 feet. Those findings suggest the sensing method may be useful outside tightly controlled laboratory conditions, although broader clinical validation would be needed before medical use.

Cost is another part of the pitch. ESP32 chips cost a few dollars, while Raspberry Pi devices cost about $35 (source). For organizations evaluating room-based monitoring across many locations, inexpensive hardware could materially change deployment economics.

Could an existing Wi-Fi environment become a passive health-monitoring network? Potentially, but the commercial path contains more than a hardware question. Healthcare buyers would likely examine validation methods, cybersecurity, consent, data handling, integration with clinical workflows and the distinction between wellness monitoring and regulated medical functionality.

The underlying model's ability to work in unfamiliar rooms may help. The lead researcher noted that Pulse-Fi generalized well in a new environment that had not been represented in its training data. "The model generalized well in [a new] setting, showing it's not just memorizing, but actually learning patterns that transfer to new situations," the researcher noted.

That said, the experiments have so far monitored only one person in a room at a time. Multiuser environments present a tougher signal-separation problem because several people may be moving, breathing and producing overlapping changes in the wireless channel. The University of California, Santa Cruz researchers are beginning pilots designed to address that limitation.

Privacy will require attention as well. Pulse-Fi does not capture facial images, but invisible environmental sensing can still raise questions about awareness and consent. Enterprises considering the technology would need clear policies explaining where monitoring occurs, who can access measurements and how long data is retained.

The lead researcher plans to establish a company to commercialize Pulse-Fi. The research group is also exploring whether the same sensing approach could monitor breathing rates or indicators associated with sleep apnea. If multiuser performance, validation and governance concerns can be addressed, Pulse-Fi could move Wi-Fi beyond connectivity and into a new role as low-cost ambient health infrastructure.