An Automated Algorithm for Estimating Respiration Rate from PPG Signals

Abstract

Objective: This paper presents a simple automated method to estimate respiration rate (RR) from the photoplethysmography (PPG) signals. Methods: The method consists of preprocessing, extremely low-frequency two-pole digital resonator, fast Fourier transform, spectral magnitude computation, prominent spectral peak identification and respiration rate computation. Validation Dataset: The proposed RR estimation method is evaluated using standard PPG databases including MIMIC-II and CapnoBase. Results: The proposed method had estimation accuracy with absolute error value (median, 25th–75th percentiles) of 0.522 (0, 1.403) and 0.2746 (0, 0.475) breaths/minute for the PPG recordings from the MIMIC-II and CapnoBase databases, respectively for the 30 s segment. Results showed that the proposed method outperforms the existing methods such as the empirical mode decomposition and principal component analysis (EMD-PCA), ensemble EMD-PCA, and improved complete EEMD with adaptive noise-PCA methods. Conclusion: Results demonstrate that the proposed method is more accurate in estimating RR with less processing time of 0.0296 s as compared to that of the existing methods. This study further demonstrate that the two-pole digital resonator with extremely low-frequency can be simple method to estimate the RR from the PPG signals.