Forecasting of heart rate variability using wrist-worn heart rate monitor based on hidden Markov model

In this paper, we present Hidden Markov Models (HMM) approach for forecasting the changes of heart rate. Heart rate is an important indicator of the state of our body. Forecasting changes of heart rate is equivalent to forecasting changes of the body state. We use numerous HMM models that is trained by datasets clustered on similarity basis. We find the optimal models with best probabilities in various learned HMM models and use this model to predict next heart rate variability. The heart rate data are collected by Fitbit-HR from 190 healthy persons. The prediction performance was accuracy = 91.87% and recall = 91.67%.