Body sensor networks based sensor fusion for cardiovascular biosignal predictions

Abstract

A Multivariate Autoregressive (MAR) model based sensor fusion technique is developed in this work to improve the reliability and accuracy of cardiovascular biosignal prediction by taking advantages of combining sensory data from heterogeneous biosensors in a network. The importance and potential applications of sensor fusion in body sensor networks are introduced. Real-world data from MIT-BIH multi-parameter database MIMIC (Multi-parameter Intelligent Monitoring for Intensive Care) is used to verify the MAR model performance. The effects of model order number and involved biosignal number are studied. The experiments also show that ECG signals can be partially recovered from other biosignals even if the ECG input is completely missing.