The Prediction and Error Correction of Physiological Sign During Exercise Using Bayesian Combined Predictor and Naive Bayesian Classifier

Abstract

Physiological signs monitored by wearable devices can reflect human body burden and exercise intensity. Due to the risk, avoidance of excessive intensity of exercise, energy-saving requirement, and other factors, it is of great necessity to predict physiological sign values for the monitoring of the human body during exercise. Most available works have used a single model for prediction of physiological signs which has a bad performance with a greater prediction error. In this light, we formalize a multistep prediction scheme for physiological signs during exercise using the Bayesian combined predictor and propose an error correction mechanism to correct the accumulated error generated in the prediction process using a naive Bayesian model. Finally, we evaluate the performance of the proposed scheme using actual monitored data of several exercisers. The simulation results show that our scheme outperforms all available schemes on the performance of prediction error.