Heartbeat time series can be described as inverse Gaussian model. The model’s time-varying parameter can be estimated by adaptive point process and then heart rate can be predicted. Based on current one-step parameter prediction

the prediction of model’s parameter mean value by incorporating weighted observation term is proposed to improve the predictive accuracy of model’s time-varying parameter. Furthermore

the point process adaptive filtering is simplified by using linear process so as to acquire real time prediction. The simulations demonstrate that our algorithm is effective in improving the efficiency of heartbeat prediction.