Low-power Continuous Heart and Respiration Rates Monitoring on Wearable Devices

Abstract

A low-power consumption approach for continuous heart rate (HR) and respiration rate (RR) monitoring for wearables is proposed. The approach provides a fusion of algorithms for HR/RR monitoring in calm and active states. For both states the algorithms process only accelerometer data. In calm state HR/RR tracking algorithm is based on a ballistocardiogram (BCG) processing. It shows 6.1 bpm root-mean-square error (RMSE) for HR monitoring and 1.9 breaths per minute RMSE for RR monitoring. In the active state HR and RR trends are calculated by machine-learning (ML) model adding a special post-processing procedure (RMSE = 7.5 bpm). The complete approach allows reducing power consumption up to 70%, compared with the state-of-the-art approach that uses a photoplethysmogram (PPG) sensor. The work results in the background for energy effective continuous monitoring of human physiological parameters by wearable devices.