Respiration Rate Estimation From Noisy Electrocardiograms Based on Modulation Spectral Analysis

Abstract

This paper presents a novel method to estimate the respiration rate (RR) from a noisy electrocardiogram (ECG) signal. The method exploits the second order periodicity of the ECG signal, caused by the influence of respiration, and relies on the so-called modulation spectral signal representation to quantify RR from the noisy ECG. The methodology is validated on two datasets, one collected at rest using medical-grade sensors and another with users wearing an off-the-shelf smartshirt throughout their workday. The paper also explores the impact of ECG recording duration on RR estimation. Results show that ECG signal recordings of 120 seconds, or longer, lead to an adequate RR estimate with an error percentage ≤12.5%.