Respiratory rate estimation from the photoplethysmogram via joint sparse signal reconstruction and spectra fusion

Respiratory rate (RR) estimation from the photoplethysmogram (PPG) is a challenging problem due to the nonstationarity of RR and disturbance. In this work, we propose a novel approach to estimate RR from the PPG signal using joint sparse signal reconstruction (JSSR) and spectra fusion (SF). A window of PPG signal is segmented into multiple overlapped measurements. Sparse spectra of these measurements are estimated by JSSR using the regularized M-FOCUSS algorithm. The kurtosis of each spectrum is used to classify it into three signal quality categories, and spectra in the highest signal quality category are fused using Respiratory Rate Tracking (RRT) to estimate RR. Validated on a public benchmark database CapnoBase, our approach outperforms a state-of-the-art algorithm in accuracy and robustness in low signal quality conditions. This is the first time JSSR has been used for RR estimation from the PPG signal. In addition, our approach works well with a low sampling frequency of 10 Hz which has great potential to be used in low-cost wearable devices.