Multi-Sensor Information Fusion And Its Applications On Wearable Computer

Wearable computer is a new type of mobile computer which is small and wearable. It can provide user with the services of information acquirement, exchange and process. It is not only the micro computer that be put on the body but also connected close to the user. Except its functions of calculation and display, using multi-channel sensors, the technology of information fusion based on wearable computer can improve user's perception of the situation about themselves and the environment. Thus it can make quick and automatic response to the user and can be more intelligently communicate with user.This paper analyzes the features of wearable computer and human physiological index. Due to the factors includes demand of real-time information fusion, features of time sequence data and the fuzziness of physiological data, this paper presents a multi-sensor information fusion method based on fuzzy dynamic Bayesian network (FDBN). Dynamic Bayesian network(DBN) can estimate the states in some time slice or even every time slice given the observation sequence, while the FDBN imports fuzzy observation that could belongs to several states at the same time and can be used to solve the problem of qualitative inference with continual observation nodes. In this paper, a systemic study on the method of modeling and inference in FDBN is given. Besides, forwards-backwards algorithm and interface algorithm for FDBN is derived by combined basic inference formula of FDBN and inference algorithms in DBN. Based on FDBN, the multi-sensor information fusion architecture and operating steps is given.Using this method, this paper builds two real-time monitoring models for cardiogenic shock and human physiological state. By online inference, it is feasible to do real-time multi-sensor fusion and assessment of the state of fusion target. The result of simulation proves that this method is practical and effective. And it could effectively eliminate the noise caused by instantaneous failure of the sensors. Besides, by analysis of probability curve, trend of human physiological state change could be acquired at the first moment.Finaly,by adopting the monitoring model builded by FDBN and using some physiological fabric sensors to get physiological data, this paper designs and implements a wearable human health monitoring system on high performance embedded system and monitoring server. In the experiments, this prototype could achieve two basic functions that are physiological data display and automatic assessment of human physiological state, and then could monitor human health remotely based on wearable computer.