SENSORFORGE

Research logger for Polar sensors

Record raw ECG, PPG, accelerometer, heart rate and RR data from Polar sensors on iPhone, as plain CSV files you own.

Get it on the App Store

The example notebooks and their data are public too, so you can try the analysis side before your first recording.

Privacy policy · Support · Example notebooks

The app

Pick a sensor, choose the streams it offers and press Start logging. While a log runs you see a live trace, the number of samples per stream and the rate each stream actually arrives at. Files are written as you record, also with the screen locked.

Record screen: a connected Polar H10 with HR and ACC selected and ECG locked, a second sensor connecting, session details and a Start logging button.
Before a log. Each connected sensor with its streams. Free: heart rate plus one raw channel; Pro: every channel and several sensors at once.
Recording screen: timer at 00:20 with two sensors, a live three-channel PPG trace, and per-stream sample counts and measured rates for a Polar Sense and a Polar H10.
While recording. Live PPG from the arm sensor, and sample counts and measured rates for every stream of both sensors.

What the data looks like

A real recording: a Polar H10 on the chest and a Polar Verity Sense on the upper arm, worn for 4.8 hours on an ordinary day out, lunch included. Every figure below was plotted straight from the raw CSV files with pandas, SciPy and Matplotlib. The notebooks and a 30-minute excerpt of this log are on GitHub.

Eight seconds of six raw streams on one time axis: ECG, filtered PPG, chest and arm acceleration, gyroscope and magnetometer, with vertical lines at the ECG R peaks.
Eight seconds, six streams, one clock. ECG and acceleration from the chest strap; PPG, acceleration, gyroscope and magnetometer from the arm sensor. The vertical lines mark ECG R peaks, so each beat can be followed across both sensors.
Time–frequency spectrogram of 4.8 hours of PPG. A bright band between 60 and 100 beats per minute traces the heart rate through the day.
The whole day as a PPG spectrogram. Raw optical signal from the upper arm, 16-second windows. The bright band is the heart rate, its second harmonic sits faintly above it, and the narrow gaps are Bluetooth dropouts, each one logged in session.json.
Fifteen-minute PPG spectrogram with the chest strap's heart rate drawn on top; the line runs along the bright band.
Zoomed in: fifteen minutes around lunch. 8-second windows, with the H10's heart rate drawn on top. The optical heart-rate band from the arm and the ECG-based heart rate from the chest follow each other.
Three panels: a Poincaré density plot of all beat pairs, a Poincaré plot of a 5-minute rest window with its SD1/SD2 ellipse, and the RR spectrum of that window split into VLF, LF and HF bands.
HRV from the H10's RR intervals. Poincaré plot of every beat pair in the recording, a quiet 5-minute window with its SD1/SD2 ellipse, and that window's spectrum split into VLF, LF and HF bands.

Example notebooks

Free, MIT-licensed Jupyter notebooks with real example data. They run on your own computer, in VS Code or in Google Colab, need only pandas, NumPy, SciPy and Matplotlib, and send your data nowhere.

Open on GitHub