A multi-step heart rate prediction method based on physical activity using Adams-Bashforth technique

Physical activity (PA) is commonly recognized to directly influence changes in heart rate (HR). HR prediction based on PA can be a useful tool in medical research and monitoring in a clinical setting. In our previous works, predictors with high accuracy were designed. However, the HR could only be predicted in single time steps. In this study, a multi-step HR prediction method is proposed. Firstly, the HR prediction problem was converted into an Initial-Value Problem for Ordinary Differential Equation (IVPODE). Then the Adams-Bashforth method was used to implement multi-step prediction. Only HR at the initial time step and PA signals are needed. Experiments were conducted based on the real-life signals from a healthy male. Predicted HR can approximately trace the actual HR in a long time step. The results show the potential of this proposed method.