Zoetude
Camera method · Research Beta
What the camera observes—and where it stops.
Zoetude uses a progressive 15–60 second protocol. Each complete window may add a bounded observation, while raw frames and signal processing remain on the device.
Transparent method · fail-soft record · no diagnostic interpretation
Direct answer
How does a Zoetude camera check-in work?
The browser tracks a face, selects stable skin regions, and analyzes small RGB changes over time. Motion, lighting, frame cadence, regional agreement, and spectral structure determine which Research Beta observations can be formed. The completed check-in remains useful even when one optional estimate is unavailable.
Progressive windows
What changes at 15, 30, 45, and 60 seconds?
Capture time is not presented as a medical accuracy guarantee. It supplies more frames and beat intervals for internal agreement checks.
First complete snapshot
Camera pulse estimate, processed rhythm shape, frame and motion context, and regional Face Response may be saved. The user can finish here.
Breathing and agreement
Consecutive pulse windows can be compared, and a camera-derived breathing estimate may be added when the respiratory channel is usable.
Beat-interval context
More detected beats can support a device-derived HRV RMSSD estimate and additional segment agreement.
Final refinement window
The last window increases the beat and respiration context available to the same gates. It does not unlock a new diagnosis or health score.
Local pipeline
What is processed on the device?
Frame
Face position, scale, head angle, motion, exposure, frame cadence, and stable facial regions are checked using the live camera stream.
Signal
Regional RGB traces are normalized, motion-sensitive segments are controlled, and pulse-band candidates are compared across regions and time windows.
Record
Only bounded results and raw-free visual summaries enter the local history. Raw frames, full landmarks, and frame-level signals are not uploaded.
Output contract
What does each displayed value mean?
| Camera pulse | A quality-gated pulse-rate estimate from camera color variation. Research Beta; not ECG, pulse oximetry, or clinical heart-rate monitoring. |
|---|---|
| Breathing | A camera-derived breathing-rate estimate from supported low-frequency signal and motion channels. Research Beta; not clinical respiratory monitoring. |
| HRV RMSSD | A device-derived estimate from processed camera beat intervals. Research Beta; not ECG HRV and not converted into stress, recovery, readiness, or disease interpretations. |
| Face Response | A raw-free regional comparison of normalized color ratios. It does not claim temperature, blood flow, perfusion, redness, sweat, inflammation, or skin condition. |
| Pulse field | A visual map of regional signal support and phase coherence. It is not an anatomical vessel map and does not show the direction of blood flow. |
Fail-soft, not fabricated
What happens when the signal is limited?
A completed check-in still preserves the context it genuinely obtained: capture duration, frame observation, user-selected state, supported channels, and a local history point. A missing physiological estimate is not replaced with a guessed number. The interface can invite another observation without turning the day into a failed ritual.
Scientific basis
What supports camera-based pulse observation?
Published research shows that consumer cameras can capture remote photoplethysmographic variation under suitable conditions. It also documents sensitivity to lighting, motion, skin appearance, camera behavior, and algorithm choice. Zoetude therefore labels its camera-derived estimates Research Beta and keeps them outside diagnosis.
- Verkruysse W, Svaasand LO, Nelson JS. Remote plethysmographic imaging using ambient light. Optics Express. 2008;16(26):21434–21445.
- de Haan G, Jeanne V. Robust pulse rate from chrominance-based rPPG. IEEE Transactions on Biomedical Engineering. 2013;60(10):2878–2886.
- Moço AV, Stuijk S, de Haan G. Motion robust PPG-imaging through color channel mapping. Biomedical Optics Express. 2016;7(5):1737–1754.
Method owner
Junhwan “Bradley” Kwon, PhD
Biomedical engineer and non-contact monitoring researcher working across anesthesiology research, thermal imaging, patient monitoring, and privacy-preserving camera interfaces.
Use the observation as context—not a verdict.
Zoetude is designed for personal wellness reflection and an optional, bounded AI conversation.