A Hybrid Model for Realistic Heart Rate Variability Simulation Based on Physiological, Stochastic, and Fractal Dynamics

Abstract

Heart rate variability (HRV) is a non-invasive biomarker reflecting the activity of the autonomic nervous system and cardiovascular regulation. The generation of realistic HRV time series can be useful in testing new signal processing algorithms, especially those based on neural networks and machine learning. In this paper, a hybrid model for HRV simulation is proposed, integrating frequency components, stochastic variability and fractal dynamics. The model allows its parameters to be adjusted, as well as to add new ones later. By changing the parameters of the model, cardiac signals with normal heart rhythm and with abnormal heart rhythm can be mathematically synthesized. The generated signals are evaluated with time, frequency and nonlinear HRV metrics. The parameters of the simulated signals are validated with internationally approved ones, as laid down in the HRV Standard. The model can generate signals of different lengths, and when generating long-term records (24-hour signals), longterm dynamics are incorporated. The results show good discrimination between time series with normal heart rhythm and those with abnormal ones. The proposed model for HRV simulation can be used in biomedical and e-health applications.