Reseach On The Distribute Intelligent Pension System Based On Data Fusion

Along with the silver hair wave attacks, the quantity of elderly increased dramatically, theirpension problems become more striking. Our special national conditions decide that old-age pensioninstitutions is the only way to solve the current pension problem, however, most of China’s currentpension agency behind the existence of facilities and equipment, shortage of staff, managementinformation system is relatively isolated, elderly health problems can not be found when so the besttreatment time delay and other issues, making the elderly can not get good care, family members arealso not assured.This paper presents a distributed intelligent pension system can effectively solve the problem of thenursing home,which not only make elderly happy but also family members can put their hearts instomach. In nursing homes, the elderly wearing the RFID smart bracelet which can capture the lifecharacteristics information, considering on the spiritual needs of the elderly, we will provideintelligent mobile robot for them so that they can chatting online with their family,meantime,thecameral can be used to collect the action information. In order to know about the physical condition ofthe elderly in real time,we need to fution two different properties information which are the lifecharacteristics information and the image motion characteristics.Data fusion technology is an effective way to get the optimal decision from the different kinds ofdata. Because of the better fault tolerance and strong adaptive ability, neural network is widely used infield of data fusion. During the Feedforward neural networks, radial basis function neural network showmore vitality than the traditional BP neural network with its fast learning speed and multi-tasking.Therfore radial basis function neural network is applied to fusion these two different types of data in themiddle layer of the implementation of distributed intelligence pension system. Due to the huge numberof raw data directly is input to the neural network may make the network structure becomes complicated,which will seriously affect the convergence speed, at the same time, a large number of redundant dataand noise data will also affect the performance of the system, so this paper use multi-tree algorithm toreduct the input data, the purpose is to improve the convergence speed and recognition precisionnetwork. Because the system data fusion requires high accuracy and fast convergence rate,however, Selectthe parameters of the traditional RBF neural network have a great impact on network performance, inorder to ensure that the system operation normally, this paper presents an improved PSO to optimizationRBF neural network,and simulation experiments using MATLAB to verify the effectiveness of theimproved algorithm.