Estimation of heart rate and respiratory rate from the seismocardiogram under resting state

Abstract

This study proposes a dedicated algorithm to estimate heart rate (HR) and respiratory rate (RR) from seismocardiogram (SCG). The proposed algorithm primarily consists of the wavelet decomposition (by db6 wavelet), the Fourier-based envelope detection and the time-averaged power spectral density (PSD) from the scalogram by complex Morlet wavelet. The records in the combined measurement of ECG, breathing and SCG (CEBS) database of PhysioNet are adopted to evaluate the performance of the proposed algorithm for HR and RR estimation. The performance was evaluated by intraclass correlation coefficient (ICC) and Bland-Altman’s agreement analysis.The statistical results for HR estimation show good to excellent correlation between the approaches by SCG and ECG and also meet the maximal allowable error for the HR meter. As the quality of some respiratory signals in the CEBS database is not good for RR estimation, additional experiments have been conducted in the university laboratory under well-controlled procedure. The statistical results for the RR estimation in these experiments show good to excellent correlation between the approaches of RR estimation by SCG and respiratory signal. The RR estimation from SCG also meets the specification given in contemporary medical device.The proposed algorithm does not take the removal of motion artifact into account and thus SCG signal must be measured under resting state. From this study, SCG can be a potential tool to estimate HR and RR for monitoring the patient in intensive care unit (ICU) or monitoring the sleep quality in clinical setting or in daily life.