Face camera
Thirty seconds of light from your face. Raw video never leaves the phone.
How Health MindTrack measures your vitals
Camera and microphone on your phone. Optical pulse (rPPG) and heart sound (PCG) stay on the device. The browser never runs the scan.
Camera, microphone, on the phone
Microphone
Heart sound on the same timeline as the optical channel. Still in development.
On-device
The scan runs on the phone. The cabinet shows numbers, not the recording.
Face · BVP
Heart sound
Live
Wearable · HR
What rPPG is
Remote photoplethysmography reads tiny colour changes in skin as blood pulses. We extract that waveform from a face recording, score quality on every attempt, and withhold the number when the signal is too weak.
What PCG is
Phonocardiography is heart sound from the microphone, aligned on the same timeline as the optical channel. The cardio channel is in development. It is not a stand-in for a clinician.
How we check ourselves
Camera and microphone are two independent signals. When they agree on heart rate, we show that agreement. When they differ, we show the difference.
Face pulse (rPPG) and heart sound (PCG) are physically independent channels. A match within 3 bpm is strong agreement; within 6 bpm is moderate. Take both within about 10 minutes to see how closely they agree. Camera-only scanners cannot make this check. The score is a cross-channel confirmation — not a diagnosis.
In the app and cabinet this is the Signal Trust Score card after a paired face scan and heart listen. See Accuracy for field numbers when we have them.
Voice as a signal
A short spoken phrase is another wellness signal — Voice Steadiness, Voice Energy, and Speech Pace — not a mental-health questionnaire.
Principle
Acoustic features of a 10-second phrase: pitch variation, voiced-segment energy, and how much of the take is speech versus pause.
On the phone
Optional after a heart listen, or from Home. Processed on the phone. Derived features can join the training corpus if you grant research. Raw voice audio is not stored unless you grant a separate voice-audio research consent. Analysis needs a separate consent from research training.
What you see
Three descriptive 0–100 scores. Same-day they are a 10% factor in Wellness Score, next to Mood check-in. Informational only — not a medical device reading.
Voice Steadiness
Less pitch and amplitude variation in voiced speech.
Voice Energy
Voiced-segment loudness and how much of the take is speech.
Speech Pace
Share of the recording that is voiced, not pause.
Same-day Voice Wellness Signature is a 10% factor in Wellness Score, next to Mood check-in. Weights are renormalized when a signal is missing — see the table below.
Breath Sound Clarity
Research participation — record a short breath take so we can train our own on-device models. This is not a scan of your breathing health.
Principle
Acoustic texture of a 12-second breath recording at arm's length: event count, spectral centroid, spectral flatness, and zero-crossing rate. Features may join the training corpus — not a health scan.
On the phone
Optional research step. Processed on the phone. Derived features can join the training corpus if you grant research. Compact breath audio is not stored unless you grant a separate breath-audio research consent.
What you see
One of three recording-quality labels: clear take, some texture noted, or quality too low. Research data only — not a medical device reading and not part of Wellness Score.
Breath sound is clear
The take had enough events and a quieter acoustic texture.
Some texture in this take — recording quality, not a health assessment
The take had more high-frequency texture. This is not a health assessment.
Recording quality too low — try again in a quieter place
Too few events, too quiet, or too noisy to describe. Try again in a quieter place.
Research data · recording quality only. Not a health scan or medical diagnosis.
Breath Sound Clarity is an experimental research feature. It collects acoustic features from a short breath recording on your phone to train our own descriptive models. It is not a diagnostic instrument, not a medical diagnosis, and not meant to assess your breathing. Do not use it if you have breathing complaints — get in-person care instead.
This consent is separate from research training, from microphone as such, and from voice analysis. Compact audio needs a separate breath-audio research consent. See Terms of Service.
Measured today
Heart rate, HRV, breathing, and stress load from a face scan. Quality is shown on every reading.
The diagrams are schematic — not recordings from users.
Heart Rate
Measured- 1
Principle
Blood under the skin slightly changes color with each beat.
- 2
On the phone
The camera watches that color change on your face for about thirty seconds. Raw video stays on the phone.
- 3
What you see
We turn the pulse waveform into beats per minute and show signal quality on every reading.
From a face scan. Field performance is typically lower than lab sets. Quality is shown on the reading instead of a headline percentage. Accuracy
HRV
Measured- 1
Principle
The time between beats is not perfectly even — that variation is heart-rate variability.
- 2
On the phone
The same face scan yields intervals between beats, only when quality is high enough.
- 3
What you see
We report HRV in milliseconds against your own usual, not a population cut-off.
From a face scan. Withheld when quality is too low rather than filled in with a guess. Accuracy
Breathing
Measured- 1
Principle
Face color and small motion also carry a slower breathing rhythm.
- 2
On the phone
The phone extracts that slower wave from the same camera recording.
- 3
What you see
Breaths per minute, quality-gated like heart rate.
From a face scan. An optical estimate from the face scan, not a chest strap. Accuracy
Stress
Measured- 1
Principle
Heart rhythm and breathing shift when your body is under load.
- 2
On the phone
The scan combines those signals into a 0–10 arousal level.
- 3
What you see
This is a load signal compared with your usual — not a mood label and not a clinical scale.
From a face scan. A wellness estimate. It does not name a condition. Accuracy
In development
Oxygen, blood pressure, and hemoglobin are early camera estimates. HbA1c is logged from you or a lab. We are collecting first paired data — there is no invented accuracy chart.
Oxygen
Estimate · early version- 1
Principle
Oxygen in blood slightly changes how blood reflects light.
- 2
On the phone
The camera compares how red and blue light from your skin changes with each beat.
- 3
What you see
From that ratio we estimate oxygen — still an early estimate; we are learning against a pulse oximeter.
From a face scan. Collecting first paired data. Marked as an estimate until then. Accuracy
Blood Pressure
Estimate · early version- 1
Principle
Pressure is the force of blood moving through vessels.
- 2
On the phone
A cuff or watch is the reliable source. The camera may offer a rough heuristic when nothing else is logged.
- 3
What you see
We are studying pulse-wave shape. Until paired cuff data is in, the camera number is marked as an estimate.
From a face scan. Collecting first paired data against a cuff. Prefer a logged or wearable reading when you have one. Accuracy
Hemoglobin
Estimate · early version- 1
Principle
Hemoglobin carries oxygen and slightly affects how blood absorbs light.
- 2
On the phone
That is a very thin signal the camera only sees approximately.
- 3
What you see
A lab blood test is the precise figure. We are learning a rough level by pairing scans with your lab results.
From a face scan. Collecting first paired data against lab results. Marked as an estimate. Accuracy
HbA1c
Lab / manual- 1
Principle
HbA1c reflects average sugar over the last two to three months.
- 2
On the phone
This is not an instant signal — the phone cannot see it in thirty seconds.
- 3
What you see
We show a trend from your logs and lab entries over time, not a camera reading.
Entered by you or a lab. Logged from you or a lab. No camera figure. Accuracy
How the Wellness Score is calculated
Wellness Score is a 0–100 number built from your recent readings against your baseline — heart rate, HRV, breathing, and load. It is wellness data. It is not a medical conclusion, and it does not replace a clinician.
The public ring on the home page is a demonstration on sample days. Your cabinet uses the same formula on your own scans.
Features are z-scores over 1 / 3 / 7 / 28 days, deltas, time of day, weekday, manual-entry context, signal quality, and how many channels agreed.
| Signal | Weight | Direction |
|---|---|---|
| Heart rate | 20% | Closer to your usual is better |
| HRV | 35% | Higher than your usual lifts the score |
| Stress load | 25% | Lower than your usual lifts the score |
| Breathing | 15% | Closer to your usual is better |
| Mood check-in | 10% | When logged — informal check-in, not a questionnaire |
| Voice Wellness Signature | 10% | When logged — mean of Voice Steadiness, Voice Energy, Speech Pace |
Low signal quality scales the total down (a 15% or 5% penalty below quality 0.4 / 0.6). Missing signals — including a mood check-in that has not been logged — are dropped and the remaining weights are renormalized. Until about seven days of your own data exist, the product marks the number as provisional and says so.
Accuracy and limits
Health MindTrack is a wellness tool, not a medical device. Measurements are for personal insight only.
© 2026 EDMC. All rights reserved.