Baseline Wander Removal in PPG
Baseline wander in PPG is a low-frequency drift (<0.5 Hz) caused by respiration, vasomotor activity, sensor pressure changes, and temperature variation. Removal methods include high-pass filtering, cubic spline interpolation through onset points, polynomial detrending, and morphological operators.
Baseline wander complicates amplitude-based analyses (perfusion index, pulse amplitude variability) and can cause false beat detections in peak-picking algorithms. The wander typically occupies 0.01–0.5 Hz, overlapping with clinically relevant content: respiratory modulation (0.15–0.4 Hz) and very-low-frequency HRV (0.003–0.04 Hz). Aggressive baseline removal that eliminates respiratory content is acceptable for HR extraction but unacceptable for respiratory rate estimation or frequency-domain HRV analysis.
Cubic spline interpolation through detected pulse onset (foot) points creates an estimated baseline that is subtracted from the original signal. This method preserves respiratory modulation above the baseline while removing slow drift, but requires reliable onset detection — creating a circular dependency that is resolved iteratively. Polynomial detrending (order 3–6 over 10–30 second windows) removes trends without requiring beat detection but may introduce Gibbs artifacts at window boundaries.
Morphological operators (opening followed by closing with a flat structuring element wider than one pulse cycle) estimate the baseline by extracting the signal envelope, providing robust baseline estimation without frequency-domain assumptions. This approach is effective for irregular rhythms (AF) where spline-based methods may fail due to inconsistent onset detection.
Frequently Asked Questions
What high-pass cutoff frequency removes baseline wander without affecting PPG?
0.5 Hz removes most baseline wander while preserving cardiac content. 0.1 Hz preserves respiratory modulation. 0.05 Hz preserves VLF HRV. The appropriate cutoff depends on which low-frequency components are needed for analysis.
Can baseline wander affect heart rate accuracy?
Severe baseline wander can cause missed or false peak detections, affecting HR accuracy by 1–5 bpm. Template matching and derivative-based peak detection are more robust to baseline wander than amplitude threshold methods.
How do consumer wearables handle baseline wander?
Most wearables apply fixed high-pass filtering (0.5–1 Hz cutoff) as the simplest solution. Research-grade devices may use adaptive methods. The high-pass approach sacrifices respiratory and VLF information for reliable cardiac signal extraction.