基于BP神经网络的数据融合技术在分布式养老系统中的应用

根据神经网络在数据融合的应用比较成熟,BP神经网络具有实现简单,以及在一定范围内具有较高识别精度的特点,选用此方法,在分布式养老系统中对老人的体征信息和监控设备两种属性不同的图像信息进行融合处理,实时监测老年人的身体状态.当神经网络的输入信息维数过高时,会导致神经网络训练速度下降.针对此问题,对传统的基于神经网络的融合算法进行改进,利用粗糙集对输入数据进行约简,使神经网络输入数据降维.同时,将约简后的信息进行训练.算法在训练时间和融合结果的准确性上都有提高.

Application of Data Fusion Technique in Distributed Pension System based on BP Neural Network

In order to monitor the elderly＇s physical state in real time, the signs information and image information in the distributed pension system need to be fused. BP neural network algorithm was chosen for it is simply used and has the characteristics of high recognition accuracy in some extent. When the neural network＇s input dimension is too high, the neural network training rate declines. Rough set was adopted to contract the input data in order to improve the traditional data fusion algorithm based on neural network, which decreases the neural network input data dimension. At the same time, the contracts information was trained, which means both of the training time and the accuracy of the results have improved.