Noncontact Monitoring of Heart Rate Variability Using a Fiber Optic Sensor

Abstract

Heart rate variability (HRV) is widely investigated to provide early warning signs for cardiovascular diseases (CVDs). However, traditional HRV monitoring methods are inconvenient in daily long-term continuous monitoring due to the uncomfortable direct contact of the devices with the skin. In this work, we present a ballistocardiogram (BCG)-based system for long-term HRV monitoring in a non-contact mode. BCG signals can reflect the mechanical activity of the heart, from which the beat-to-beat interval (BBI) can be extracted for HRV analysis. An effective solution is demonstrated for accurate and steady BCG signal detection using a highly-sensitive fiber optic sensor. A robust heartbeat extraction algorithm based on the deep learning method is proposed to improve the accuracy of the BBI extraction. Continuous wavelet transform (CWT) is introduced to facilitate feature extraction against the variation of morphological characteristics of BCG signal. The accuracy of BBI estimation presents a median error of 4.4 ms, indicating the effectiveness and feasibility of the proposed system. This work is expected to pave a new and practical pathway for household HRV monitoring during sleep.