Deep Recurrent Neural Network for Extracting Pulse Rate Variability from Photoplethysmography During Strenuous Physical Exercise

Abstract

Pulse rate variability (PRV) extracted from photoplethysmography (PPG) signal is a promising surrogate for heart rate variability (HRV) and has shown its great potential in diagnosing cardiac dysfunctions and autonomic nervous system diseases. However, the accurate extraction of PRV during strenuous physical exercise faces enormous challenges due to PPG’s extreme vulnerability to motion artifacts. In this work, we introduce a deep recurrent neural network (RNN) based on bidirectional Long-Short Term Memory Network (biLSTM) for accurate PPG cardiac period segmentation. After that, three important indexes for PRV are calculated, which are peak intervals, pulse intervals, and instantaneous heart rates (IHR). Comparison results with state-of-the-art methods on a dataset including 48 subjects show the promising performance of the proposed algorithm in PRV indexes estimation and recovery. To our best knowledge, this is the first time a deep learning-based algorithm been involved for extraction of PRV from seriously corrupted PPG signals.