一种基于深度学习模型的数据融合处理算法  

Data Aggregation Algorithm Based on the Model of Deep Learning

作　　者：马永军[1] 薛永浩 刘洋[1] 李亚军[1] MA Yongjun;XUE Yonghao;LIU Yang;LI Yajun(College of Computer Science and Information Engineering,Tianjin University of Science & Technology,Tianjin 300457,China)

机构地区：[1]天津科技大学计算机科学与信息工程学院,天津300457

出　　处：《天津科技大学学报》2017年第4期71-74,78,共5页Journal of Tianjin University of Science & Technology

基　　金：天津市教委重大项目(2014ZD22)

摘　　要：针对无线传感器网络传统数据融合算法效率较低、处理高维数据困难问题,提出一种基于深度学习模型的卷积神经网络结构实现数据融合的算法CNNMDA.算法首先在汇聚节点对构建的特征提取模型CNNM进行训练,然后各终端节点通过CNNM提取原始数据特征,最后向汇聚节点发送融合后的数据,从而减少数据传输量,延长网络寿命.仿真实验表明,CNNMDA与同类融合算法相比,在同样数据量的情况下能够大幅降低网络能耗,并有效提升了数据融合效率与准确度.Traditional methods of data fusion in wireless sensor networks(WSNs)are inefficient and not ideal for processing high-dimensional data.Therefore,a data aggregation algorithm CNNMDA(convolutional neural networks model data aggregation)was proposed,which combined convolutional neural networks(CNN)and WSNs clustering routing protocol.A feature extraction model(CNNM)is designed by using CNNMDA firstly and then trained in Sink node.After that the cluster nodes use CNNM to extract data features,which are sent to the Sink node by cluster heads,thereby reducing the quantity of data transmission and extend the network lifetime.Simulation results show that compared with existing similar algorithms,the energy consumption of CNNMDA decreases obviously and the accuracy of the data fusion can be effectively improved.

关 键 词：数据融合 深度学习 无线传感器网络 卷积神经网络 

分 类 号：TP393[自动化与计算机技术—计算机应用技术] TP183[自动化与计算机技术—计算机科学与技术]