Researchers have shown that everyday Wi‑Fi can double as a contactless heart rate sensor—no chest straps, wearables, or electrodes required. Using a Raspberry Pi 4B and clever signal processing, the team demonstrated clinical‑level accuracy by extracting a pulse from the subtle ripples a beating heart imprints on wireless signals.
How Wi‑Fi becomes a stethoscope
When Wi‑Fi waves bounce around a room, they pick up tiny fluctuations from moving objects. A human chest rising, falling, and subtly pulsing with each heartbeat changes the multipath reflections that reach a receiver. Those changes appear in Channel State Information (CSI)—a fine‑grained snapshot embedded in modern 802.11 transmissions that describes how each subcarrier traveled from transmitter to receiver.
By isolating the components associated with micromotions in the 0.8–2.17 Hz band (roughly 48–130 beats per minute), the researchers turned ambient Wi‑Fi into a non‑intrusive biosensor. Unlike camera‑based solutions, Wi‑Fi works in the dark and through some obstructions, and unlike wearables, it doesn’t require user compliance once installed.
The hardware: Raspberry Pi vs. ESP32
The group validated their approach with two off‑the‑shelf platforms: a pair of ESP32 modules and a Raspberry Pi 4B. Both can expose CSI, but the Pi’s Wi‑Fi chipset samples many more subcarriers—234 versus roughly 64 on the ESP32 when operating with broader channel widths—providing richer data to reconstruct the heart signal.
The results were striking. Using ESP32 units, the system achieved 99.38% accuracy with an average error of just 0.51 beats per minute against reference measurements. With the Raspberry Pi, accuracy climbed to 99.81% and error fell to about 0.2 BPM—levels consistent with clinical instruments—using as little as five seconds of data.
Those figures align with broader literature on RF sensing. Prior academic work from labs at the University of California and groups publishing through IEEE has shown that higher subcarrier density and stable radio chains can markedly improve vital sign extraction, especially in cluttered environments.
From CSI to BPM: the signal pipeline
Turning raw Wi‑Fi traffic into a pulse reading takes several steps. First, the receiver logs CSI frames continuously while a transmitter (another Wi‑Fi device or the same Pi) emits packets. Next, the system isolates the subject’s micro‑motion by removing static paths and slow body movements, then applies a band‑pass filter tuned to heart rate frequencies.
Residual noise is reduced with smoothing and denoising techniques so peaks align with the cardiac cycle rather than respiration or random jitter. Finally, a lightweight Long Short‑Term Memory (LSTM) neural network maps the cleaned waveform to a heart rate estimate. The model size is small enough to run on the Raspberry Pi without a GPU, and inference latency is low, enabling near‑real‑time readouts.
Ground truth comparisons are essential. In the study, Wi‑Fi‑derived rates were matched against simultaneous readings from conventional sensors, a standard practice in biomedical signal validation recommended by organizations such as the Association for the Advancement of Medical Instrumentation.
Replicating the setup responsibly
A practical proof‑of‑concept needs little more than a Raspberry Pi 4B with built‑in Wi‑Fi, a secondary Wi‑Fi device to generate traffic if needed, and software to capture CSI. Open research tools like Nexmon (developed by SEEMOO Lab collaborators) can enable CSI extraction on certain Broadcom chipsets, while Espressif provides CSI examples for the ESP32 ecosystem.
Place the transmitter and receiver a few meters apart, with the subject seated between or near the line of sight. Collect five to ten seconds of CSI while minimizing large movements. Apply pre‑processing and filtering in Python or MATLAB, then run an LSTM or even a classical peak‑detection baseline to benchmark performance. Always validate against a reference—an ECG or finger pulse oximeter—to quantify error.
Two cautions: motion and multipath matter. Walking, arm swings, or multiple people nearby can overwhelm the heart signal; multi‑person separation typically requires beamforming or advanced models. And privacy is paramount. RF sensing can reveal sensitive physiological information, so consent and clear disclosure are non‑negotiable, especially in workplaces or shared spaces.
Why this matters
Contactless monitoring could complement wearables in hospitals, senior care, and sleep labs, where stray cables and skin contact are pain points. It also hints at low‑cost telehealth: a $35–$50 microcomputer running a background service could spot tachycardia episodes or alert caregivers without the user remembering to charge a device.
There are limits. Wi‑Fi can’t yet replace ECG for arrhythmia diagnosis, and performance varies by room geometry, interference, and subject position. But with commodity radios, modern 802.11 chipsets, and small neural networks, the signal is already strong enough to be useful. The broader trend—using ubiquitous infrastructure as a health sensor—echoes earlier academic projects that tracked respiration and gait via RF, and it’s moving steadily from lab to living room.
The takeaway: if your router is on, you already have part of a heart monitor. A Raspberry Pi and the right algorithms can do the rest.