Photoplethysmography-Based Heart Action Monitoring Using a Growing Multilayer Network

Abstract

Electrocardiogram (ECG) is commonly considered as the reference standard signal to measure the biopotential by the electrical signals that control the expansion and contraction of the heart chambers. However, for long-term heart action monitoring, devices used to collect continuous ECG signals under free-living conditions typically have several operational difficulties. As an alternative, photoplethysmography (PPG) signals can be collected by pulse sensors to sense the rate of blood flow as controlled by the heart’s pumping action. Consequently, ECG signals and PPG signals are intrinsically correlated to reflect the status of the heart. In this article, we study the relation between ECG and PPG by proposing a novel growing multilayer network (GMLN). The approach can design the number and topology of layers automatically. Relying on a growth mapping of input data, with the representation from the lower level to the higher level, the network is to fulfill a reconstruction implicitly to the target signal. Experimental results obtained from benchmark datasets show that the proposed method can achieve higher similarity for the waveform and more accurate heart rate (HR) detection with the target signal.