Periodic motion artifacts affect photoplethysmography (PPG) signals in activities of daily living (ADL), cardiopulmonary exercise testing (CPX), and cardiopulmonary resuscitation (CPR). This hampers measurement of interbeat intervals (IBIs) and oxygen saturation (SpO 2 ). Our objective was to develop a generic algorithm to remove periodic motion artifacts, recovering artifact-reduced PPG signals for beat-to-beat analysis.
Methods: The algorithm was retrospectively evaluated on forehead PPG signals measured while walking on a treadmill. The step rate was tracked in a motion reference signal via a second-order generalized integrator with a frequency-locked loop. Two reference signals were compared: sensor motion relative to the skin ( ∆x[n]) measured via self-mixing interferometry and head motion ( av[n] ) measured via accelerometry. The step rate was used in a quadrature harmonic model to estimate the artifacts. Quadrature components need only two coefficients per frequency leading to a short filter and prevent undesired frequency-shifted components in the artifact estimate. Subtracting the estimate from the measured signal reduced the artifacts.
Results: Compared to ∆x[n] , av[n] had a better signal-to-noise ratio and more consistently contained a component at the step rate. Artifact reduction was effective for distinct step rate and pulse rate, since the artifact-reduced signals provided more stable IBI and SpO 2 measurements.
Conclusion: Accelerometry provided a more reliable motion reference signal. The proposed algorithm can be of significance for monitoring in ADL, CPX, or CPR, by providing artifact-reduced PPG signals for improved IBI and SpO 2 measurements during periodic motion.