A UKF-Based Emergency Aware Fusion Model in a Heterogeneous Network for Wireless Body Networks

Wearable body networks (WBNs) have revealed plenty of effective technologies in many application domains, including healthcare, fitness, and smart cities. Wearable devices are integrated with some daily necessities and they gather information, such as motion data and physiological data. In general, the Kalman-based filters are used to remove the noise from the interference of these data to increase their accuracy. On the other hand, the data could also change dramatically in cases of sudden change, such as the wearer beginning to run without any warning, heart rate rises quickly when the wearer is startled, etc. The existing Kalman-based filters hardly distinguishes both noise and sudden changes. To address this problem, an unscented Kalman filter (UKF)-based emergency-aware fusion model is proposed in this paper to efficiently increase the accuracy of the collecting data when an emergency happens. Here, an early warning mechanism (EWM) is produced to discover the emergency. The traditional UKF is used when an emergency does not exist. The gain matrix should be compensated to reduce the filtering error when an emergency exists. Therefore, the estimator should be adjusted accordingly to avoid it. In simulations, important parameters are discussed. Compared with a traditional UKF, the proposed model shows the same performance without the emergency and shows even better results when an emergency happens.