A computationally efficient algorithm for estimating respiratory rate from seismocardiogram

Abstract

Respiratory rate (RR) monitoring is crucial in clinical and healthcare settings. With advancements in microelectromechanical system (MEMS) technology, monitoring respiration using seismocardiogram (SCG) has emerged as a promising approach. This study presents a computationally efficient algorithm for estimating RR from SCG signal. The proposed algorithm consists of two primary steps: first, applying Gaussian averaging filter for pre-processing, followed by computing the averaged spectrum from the scalogram based on complex Morlet wavelet within the specified frequency range. RR is then estimated from the dominant frequency that appears in the averaged spectrum. This study also establishes the analytical relationship between the parameters of Gaussian averaging filter and the sampling frequency of SCG signal, which serves as the design criterion for pre-processing filter. The proposed algorithm undergoes testing on two SCG datasets and demonstrates its feasibility for estimating RR from SCG signal. The algorithm is also tested on both finger and wrist PPG datasets for RR estimation. The test results present high agreement between the RR estimations from SCG (or PPG) signal and those from respiration signal. Furthermore, an experiment using the dataset collected with a magnetic field-based respiration sensor during pulmonary rehabilitation is conducted to validate the algorithm’s robustness against motion artifacts. Based on the testing results, the proposed algorithm shows significant potential in clinical and healthcare applications.