How do smart bands track heart rate and activity metrics?

19 Eki 11:42

The optical HR sensor (usually a green‑LED PPG photodiode) samples the blood‑volume changes at around 25–100 Hz, while the 3‑axis accelerometer runs at a similar rate to capture motion. In my own tinkering with a Fitbit‑style board, I first low‑pass filter the PPG signal (≈4 Hz cutoff) to get rid of the high‑frequency LED flicker, then apply a band‑pass (0.5–4 Hz) to isolate the pulse component. The accelerometer data is high‑pass filtered (≈0.3 Hz) to extract the dynamic movement, and then a simple adaptive filter (e.g., LMS) subtracts the motion‑induced artifacts from the PPG trace. After that, I run a peak‑detection algorithm on the cleaned PPG to compute inter‑beat intervals, which are averaged into BPM and fed into a rolling‑window HRV estimate. For activity metrics, I integrate the accelerometer’s magnitude vector, apply a step‑detection threshold, and classify intensity by the variance of the signal over 1‑second windows. All the derived numbers (HR, steps, activity level) are packed into a compact BLE payload—usually a 4‑byte header followed by two‑byte BPM, two‑byte step count delta, and a single byte activity flag—then sent at the configured connection interval (often 100 ms). This keeps the bandwidth under 10 kbps and battery usage low. If you need a quick reference, check out the Texas Instruments “HRV‑Sensor Fusion” app note (TI‑SENS‑2023) and the Nordic “Bluetooth Low Energy – Data Throughput” guide; they walk through the exact filter coefficients and packet format I used.