| 摘要: |
| 针对径向基函数（Radial Basis Function,RBF）神经网络算法在无线网络室内定位中拓扑结构和网络参数难以确定，其定位效果不理想的问题，提出了一种用核主成分分析的模糊C均值聚类算法（Fuzzy C-Means clustering algorithm based on Kernel Principal Component Analysis,KPCA-FCM）和模拟退火自适应遗传算法（Simulated Annealing adaptive Genetic Algorithm,SAGA）优化RBF神经网络的无线室内定位算法。首先利用KPCA对原始训练数据样本进行数据预处理，再通过KPCA-FCM算法计算出最优聚类数目和聚类中心点；其次将聚类数目和聚类中心点作为隐含层神经元个数和中心值，创建RBF神经网络，并将其网络参数映射到SAGA算法中；再次由SAGA算法进行网络参数寻优，把最优的解映射回RBF神经网络；最后利用样本数据对RBF神经网络进行训练和测试，完成建立RBF神经网络算法模型。实验表明，在相同的环境中，所提算法比传统RBF神经网络定位精度提高了48.41%。 |
| 关键词: 室内无线定位 RBF神经网络 核主成分分析 模糊C均值聚类 模拟退火自适应遗传算法 |
| DOI： |
|
| 基金项目:贵州省科技厅联合基金资助项目（黔科合J字［2014］2082；黔科合LH字［2016］7069） |
|
| Wireless Indoor Positioning Based on Optimized RBF Neural Network |
| LIU Xia,MO Shupei,HE Huiling,YANG Jun |
| (1.School of Electronics and Information Engineering，Guizhou Industry Polytechnic College，Guiyang 551400，China;
2.School of Automation，Central South University，Changsha 410083，China) |
| Abstract: |
| It is difficult to determine the topological structure and network parameters in indoor positioning of wireless network by means of radial basis function(RBF) neural network algorithm,and the positioning effect is not perfect.Therefore,a fuzzy C-means clustering algorithm based on kernel principal component analysis(KPCA-FCM) and simulated annealing adaptive genetic algorithm(SAGA) is proposed to optimize the wireless indoor positioning algorithm of RBF neural network.Firstly,KPCA is used to pre-process the original training data samples,and then KPCA-FCM algorithm is used to calculate the optimized clustering number and clustering centers.Secondly,by taking the clustering number and clustering centers as the number and center value of hidden layer neurons,the RBF neural network is created and its network parameters are mapped to SAGA algorithm.Thirdly,SAGA algorithm is used to optimize network parameters,and the optimal solution is mapped back to RBF neural network.Finally,the sample data is used to train and test the RBF neural network,and the algorithm model of RBF neural network is established.Experiments show that in the same environment,the positioning accuracy of the proposed algorithm is increased by 48.41%,higher than that of the traditional RBF neural network. |
| Key words: wireless indoor positioning RBF neural network kernel principal component analysis fuzzy C-mean clustering simulated annealing adaptive genetic algorithm |
《电讯技术》编辑部 