Sample and restoring algorithm of physiological signals from wireless body area network

Abstract

This paper proposes a sample and restoring algorithm of periodic time series from wireless body area network (WBAN). The high frequency sampling of sensors results in huge consumption of storage resources of server hardware. Firstly, ZigBee star network model is designed as the basic network structure of the WBAN system. Next, the dynamic storage algorithm is used to map the periodic signals of time series to the new storage samples on the server side, and the regression models of artificial neural network (ANN), convolutional neural network (CNN) and residual network (ResNet) are used to restore the samples to the original signals. The simulation results show that under the premise of using the CNN-ResNet model, the dynamic restoring algorithm has higher regression accuracy than the static storage algorithm. Finally, in order to evaluate which restoring method is the best of the dynamic storage method, a cost function is designed based on the time complexity of the algorithm, the number of storage points of periodic pulse wave and the recovery accuracy. The simulation results show that among the dynamic storage method, the cost of CNN-ResNet model is the lowest.