How do heart rate and activity sensors in smart bands work?

25 Tem 19:37

Hey man, when I got a new smart band last year and started testing it out, my first thought was, "How accurate is this thing at measuring my heart rate?" My model uses an optical HR sensor that works by detecting the light reflected off the blood vessels under the skin; the LEDs emit light at around 660 nm, sending it into the vessels, and photodiodes pick up the reflected signal. The device then runs this signal through a PPG (photoplethysmography) algorithm that filters out the sine waves to determine the pulse period. But when my wrist was moving a lot—like when I was running or shaking my arm—the signal got noisy, and the measurements would drift. The reason? The LED light was scattering due to the movement on the surface as it tried to penetrate the skin. Following the manufacturer's advice to keep my wrist steady and increasing the measurement interval from 1 second to 5 seconds (basically taking a longer average) helped improve accuracy by reducing that "low-frequency noise." I had a similar experience with the accelerometer too. The motion sensor tracks movement across three axes (X, Y, Z) to count steps, gauge activity levels, and even determine sleep stages. The position of the wrist was key here; the sensor had to sit perfectly centered and flat for the signal to calibrate correctly. Instead of just using a simple threshold, the algorithms applied dynamic filtering (like a Kalman filter) to distinguish real activity from sudden movements—like shaking the phone—from actual activity. In research, techniques like "adaptive filtering" and "machine-learning-based artifact detection" are popular for reducing these kinds of artifacts. Some studies even use real-time CNNs to classify signal quality and automatically discard bad measurements. Ultimately, the combination of sensor placement, measurement intervals, and smart algorithms really boosts accuracy, but in extreme conditions—like cold weather or sweating—some errors are still unavoidable.

25 Tem 21:12

Smartwatch optical heart rate sensors rely on photoplethysmography (PPG): a (usually green) LED shines on the skin and a photodiode measures the reflection changes caused by blood flow. Chest straps, in contrast, use electrocardiography (ECG) to directly capture the heart’s electrical signal, typically delivering higher accuracy (±1 bpm) but at the cost of comfort and the need for wired connections. The built-in 3-axis MEMS accelerometers track movement for step counting, sleep detection, and PPG signal correction (e.g., filtering out motion artifacts). Sampling rate is critical: too slow (e.g., 1 s) blurs R-R peak resolution, while too fast drains battery and amplifies noise. Wrist placement matters too—underside offers better skin contact, while the top is more prone to motion and pressure shifts. Modern algorithms (adaptive filtering, machine-learning models) fuse PPG and accelerometer data to identify and remove motion, temperature, and ambient noise artifacts. Current research includes multichannel setups—using multiple wavelengths (green, red, infrared) to offset skin tone and perfusion differences—and “PPG+ECG hybrid” sensors with embedded photodiodes that switch to ECG mode when PPG signals degrade. These advances are narrowing the accuracy gap between smartwatches and medical-grade devices while preserving the autonomy and comfort that drive their popularity.

25 Tem 22:33

I noticed while testing my first smart band that PPG measurements during exercise only provide stable values after a few seconds because skin contact pressure on the wrist and arm movement create artifacts—a slight gap between the sensor and skin drastically reduces accuracy. By adjusting the measurement interval to 1 second during running and testing algorithms that filter out sudden acceleration spikes, I was able to significantly stabilize the heart rate display.