Your phone guesses your heart rate wrong during walking or running because it discards not-quite-perfect signal data.

Measuring heart rate on smartphones using photoplethysmography (PPG), an optical sensor method, is often unreliable when people move around. Physical movements mess up the light signals differently depending on the activity, and typical systems simply throw away noisy data instead of rescuing useful parts. To fix this, researchers built MoWaveQFormer, a model that adapts to user movement. First, the system checks movement data from an accelerometer to figure out the activity type. Second, it uses that activity type to choose one of three specialized filters to clean up the optical signal. Third, it breaks the cleaned signal into patches, uses a soft gate to weigh them by signal quality, and feeds them into a Transformer to calculate heart rate. In testing, the model achieved an error of 7.85 bpm, beating five other methods and running fast enough for real-time use.

Why it matters

Current smartphone heart-rate apps become unreliable the moment you move. This method stops throwing away partially useful signal, and instead tailors the processing to your exact motion. If it works on phones, millions of people with heart conditions or fitness goals would get continuous, trustworthy readings during real life, not just while lying still.

Questions answered