Physiological state assessment and prediction based on multi-sensor fusion in body area network

Abstract

In recent years, with the rapid development of biosensor technology and Body Area Network technology, more and more Body Area Network products for sensing physiological states are entering people’s lives. It has become realistic to collect and record daily human physiological data such as pulse, respiration, and body temperature on a Body Area Network device. However, Body Area Network devices collect large amounts of physiological signal data over long periods. The explosive growth of data is not conducive to the observation, analysis, and extraction of valuable physiological information by ordinary users. Therefore, this paper uses data fusion methods to calculate physiological states from sensor data. First, physiological states were categorized using different physiological signals to obtain the change of physiological states in a continuous-time. Second, the classification results of multiple physiological signals were coded and fused to obtain the distribution of different physiological state levels over a while. Finally, the weighted Markov chain method was used to predict the level of physiological state after the fusion so that the users of the Body Area Network devices can make rational decisions based on their physiological state. The experimental results show that not only the binary coding fusion algorithm proposed in this paper has functional advantages, but also the weighted Markov chains proposed in this paper has the mean absolute percentage error less than 1.6 for predicting the fusion coding results.