Heart Rate Camera

Estimates your heart rate (BPM) from a webcam.

1. Finds your face and forehead.
2. Reads small color changes in the skin (blood pulse).
3. Filters the signal and shows BPM.

Not a medical device. For learning and experiments only.

Run (Python)

python -m venv .venv
.venv\Scripts\activate
pip install -r requirements.txt
python scripts/live.py

First run downloads the face model into models/.

Keys: q quit, m motion, b breath-hold, o cover face, d dark, t bright light

Optional:

python scripts/live.py --reference-bpm 72
python scripts/live.py --signal-method green

Run (browser)

cd web
python -m http.server 8080

Open http://localhost:8080. Needs localhost or HTTPS for camera access. Nothing is uploaded.

How it works

• Face: MediaPipe face mesh, forehead region.
• Signal: RGB from forehead. Python uses CHROM by default; browser uses green channel.
• Rate: Bandpass filter (0.75-2.5 Hz), FFT peak, convert to BPM.

Settings: config/pipeline.json

Study (optional)

Record sessions and compare to a pulse oximeter:

python scripts/record_study.py
python scripts/analyze_study.py

See docs/STUDY_PROTOCOL.md.

Limits

• Sit still, normal indoor light works best.
• Motion, dark rooms, and bright torch break the reading.
• Not as accurate as a finger sensor.

Code

src/rppg/       library
scripts/        live demo and study tools
web/            browser demo
config/         settings

Private notes go in internal/ (not in git). See templates/internal/README.md.

Uses

Python, OpenCV, MediaPipe, SciPy, Matplotlib, JavaScript