Harmonic-Selective Gaussian Filtering for Morphology and Timing Preservation in PPG Signals.

Photoplethysmography (PPG) is a widely used non-invasive optical technique for assessing cardiovascular dynamics and related hemodynamic processes. However, conventional noise-reduction methods can alter signal timing and distort waveform morphology, thereby affecting the identification of physiologically relevant events. Here, we propose a frequency-domain Gaussian filtering framework for selectively extracting harmonics from PPG signals. The method combines a Gaussian band-reject filter centered at 0 Hz to suppress the dominant DC component, reduce the non-pulsatile baseline, and attenuate low-frequency contributions associated with slow modulation processes. Symmetric Gaussian bandpass filters are then applied to isolate harmonic components, with the number of retained bands adapted to the requirements of a given application. As a proof of concept, the framework was applied to both a simulated PPG waveform and an experimental PPG recording. Because of the symmetric zero-phase properties of the filters, the temporal alignment of key PPG events, including the systolic peak, diastolic decay, and dicrotic notch, can be preserved while phase distortion is avoided. Reconstruction from filtered harmonics further suggests that low-order harmonics retain much of the observable waveform structure in the signals analyzed. Overall, this harmonic-based Gaussian filtering provides a promising framework for analysis of PPG signals and motivates further investigation of its potential for extracting physiologically related information from harmonic components, such as bedside monitoring.