Reconstruction of Corrupted Photoplethysmography Signals to Facilitate Continuous Monitoring.

Photoplethysmography (PPG) is the leading technology behind wearables, significantly hindered by PPG's susceptibility to motion artifacts (MAs). This study presents a PPG reconstruction algorithm operating regardless of the source of noise. Artifact detection is initially performed by spectral and amplitude-based control. Morphology, spectral and heart-rate (HR) variability (HRV) information of the adjacent clean segments are used for reconstruction. Thirty-six originally clean PPGs with added noise of 2–120 s were used. Recordings were resampled to 250 Hz. HR and HRV features were compared between original and reconstructed PPGs via Pearson correlation ( $\rho$ ), Bland-Altman (BA), normalized root mean square error (nRMSE) and mean absolute percentage error (MAPE) analysis. For HR, $\rho > 0.9$ for all noise lengths. Time-domain HRV: $\rho > 0.91$ . Frequency-domain HRV: $\rho > 0.75$ . Poincaré indices: $\rho > 0.85$ . Max. nRMSE: 0.58%, max. MAPE: 1.72%. At least 86% of recordings were within confidence interval, with most results being between 89%–97%, regardless of the noise duration. The proposed algorithm allows the reconstruction of corrupted PPG signals. The noise duration cut-off for optimal results is 30–45 s. Nevertheless, it can be applied in longer segments, still providing satisfactory results.